net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #include <linux/errno.h>
  42 #include <linux/types.h>
  43 #include <linux/kernel.h>
  44 #include <linux/socket.h>
  45 #include <linux/sockios.h>
  46 #include <linux/net.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #ifdef CONFIG_SYSCTL
  57 #include <linux/sysctl.h>
  58 #endif
  59 #include <linux/capability.h>
  60 #include <linux/delay.h>
  61 #include <linux/notifier.h>
  62 #include <linux/string.h>
  63
  64 #include <net/net_namespace.h>
  65 #include <net/sock.h>
  66 #include <net/snmp.h>
  67
  68 #include <net/ipv6.h>
  69 #include <net/protocol.h>
  70 #include <net/ndisc.h>
  71 #include <net/ip6_route.h>
  72 #include <net/addrconf.h>
  73 #include <net/tcp.h>
  74 #include <net/ip.h>
  75 #include <net/netlink.h>
  76 #include <net/pkt_sched.h>
  77 #include <linux/if_tunnel.h>
  78 #include <linux/rtnetlink.h>
  79
  80 #ifdef CONFIG_IPV6_PRIVACY
  81 #include <linux/random.h>
  82 #endif
  83
  84 #include <asm/uaccess.h>
  85 #include <asm/unaligned.h>
  86
  87 #include <linux/proc_fs.h>
  88 #include <linux/seq_file.h>
  89
  90 /* Set to 3 to get tracing... */
  91 #define ACONF_DEBUG 2
  92
  93 #if ACONF_DEBUG >= 3
  94 #define ADBG(x) printk x
  95 #else
  96 #define ADBG(x)
  97 #endif
  98
  99 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 100 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
 101
 102 #ifdef CONFIG_SYSCTL
 103 static void addrconf_sysctl_register(struct inet6_dev *idev);
 104 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 105 #else
 106 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 107 {
 108 }
 109
 110 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 111 {
 112 }
 113 #endif
 114
 115 #ifdef CONFIG_IPV6_PRIVACY
 116 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 117 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 118 static void ipv6_regen_rndid(unsigned long data);
 119
 120 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
 121 #endif
 122
 123 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 124 static int ipv6_count_addresses(struct inet6_dev *idev);
 125
 126 /*
 127  *      Configured unicast address hash table
 128  */
 129 static struct inet6_ifaddr              *inet6_addr_lst[IN6_ADDR_HSIZE];
 130 static DEFINE_RWLOCK(addrconf_hash_lock);
 131
 132 static void addrconf_verify(unsigned long);
 133
 134 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 135 static DEFINE_SPINLOCK(addrconf_verify_lock);
 136
 137 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 138 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 139
 140 static void addrconf_bonding_change(struct net_device *dev,
 141                                     unsigned long event);
 142 static int addrconf_ifdown(struct net_device *dev, int how);
 143
 144 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 145 static void addrconf_dad_timer(unsigned long data);
 146 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 147 static void addrconf_dad_run(struct inet6_dev *idev);
 148 static void addrconf_rs_timer(unsigned long data);
 149 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 150 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 151
 152 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 153                                 struct prefix_info *pinfo);
 154 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 155                               struct net_device *dev);
 156
 157 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 158
 159 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 160         .forwarding             = 0,
 161         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 162         .mtu6                   = IPV6_MIN_MTU,
 163         .accept_ra              = 1,
 164         .accept_redirects       = 1,
 165         .autoconf               = 1,
 166         .force_mld_version      = 0,
 167         .dad_transmits          = 1,
 168         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 169         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 170         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 171 #ifdef CONFIG_IPV6_PRIVACY
 172         .use_tempaddr           = 0,
 173         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 174         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 175         .regen_max_retry        = REGEN_MAX_RETRY,
 176         .max_desync_factor      = MAX_DESYNC_FACTOR,
 177 #endif
 178         .max_addresses          = IPV6_MAX_ADDRESSES,
 179         .accept_ra_defrtr       = 1,
 180         .accept_ra_pinfo        = 1,
 181 #ifdef CONFIG_IPV6_ROUTER_PREF
 182         .accept_ra_rtr_pref     = 1,
 183         .rtr_probe_interval     = 60 * HZ,
 184 #ifdef CONFIG_IPV6_ROUTE_INFO
 185         .accept_ra_rt_info_max_plen = 0,
 186 #endif
 187 #endif
 188         .proxy_ndp              = 0,
 189         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 190         .disable_ipv6           = 0,
 191         .accept_dad             = 1,
 192 };
 193
 194 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 195         .forwarding             = 0,
 196         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 197         .mtu6                   = IPV6_MIN_MTU,
 198         .accept_ra              = 1,
 199         .accept_redirects       = 1,
 200         .autoconf               = 1,
 201         .dad_transmits          = 1,
 202         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 203         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 204         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 205 #ifdef CONFIG_IPV6_PRIVACY
 206         .use_tempaddr           = 0,
 207         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 208         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 209         .regen_max_retry        = REGEN_MAX_RETRY,
 210         .max_desync_factor      = MAX_DESYNC_FACTOR,
 211 #endif
 212         .max_addresses          = IPV6_MAX_ADDRESSES,
 213         .accept_ra_defrtr       = 1,
 214         .accept_ra_pinfo        = 1,
 215 #ifdef CONFIG_IPV6_ROUTER_PREF
 216         .accept_ra_rtr_pref     = 1,
 217         .rtr_probe_interval     = 60 * HZ,
 218 #ifdef CONFIG_IPV6_ROUTE_INFO
 219         .accept_ra_rt_info_max_plen = 0,
 220 #endif
 221 #endif
 222         .proxy_ndp              = 0,
 223         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 224         .disable_ipv6           = 0,
 225         .accept_dad             = 1,
 226 };
 227
 228 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 229 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 230 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 231 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 232 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 233
 234 /* Check if a valid qdisc is available */
 235 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 236 {
 237         return !qdisc_tx_is_noop(dev);
 238 }
 239
 240 /* Check if a route is valid prefix route */
 241 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 242 {
 243         return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
 244 }
 245
 246 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 247 {
 248         if (del_timer(&ifp->timer))
 249                 __in6_ifa_put(ifp);
 250 }
 251
 252 enum addrconf_timer_t
 253 {
 254         AC_NONE,
 255         AC_DAD,
 256         AC_RS,
 257 };
 258
 259 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 260                                enum addrconf_timer_t what,
 261                                unsigned long when)
 262 {
 263         if (!del_timer(&ifp->timer))
 264                 in6_ifa_hold(ifp);
 265
 266         switch (what) {
 267         case AC_DAD:
 268                 ifp->timer.function = addrconf_dad_timer;
 269                 break;
 270         case AC_RS:
 271                 ifp->timer.function = addrconf_rs_timer;
 272                 break;
 273         default:;
 274         }
 275         ifp->timer.expires = jiffies + when;
 276         add_timer(&ifp->timer);
 277 }
 278
 279 static int snmp6_alloc_dev(struct inet6_dev *idev)
 280 {
 281         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 282                           sizeof(struct ipstats_mib)) < 0)
 283                 goto err_ip;
 284         if (snmp_mib_init((void __percpu **)idev->stats.icmpv6,
 285                           sizeof(struct icmpv6_mib)) < 0)
 286                 goto err_icmp;
 287         if (snmp_mib_init((void __percpu **)idev->stats.icmpv6msg,
 288                           sizeof(struct icmpv6msg_mib)) < 0)
 289                 goto err_icmpmsg;
 290
 291         return 0;
 292
 293 err_icmpmsg:
 294         snmp_mib_free((void __percpu **)idev->stats.icmpv6);
 295 err_icmp:
 296         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 297 err_ip:
 298         return -ENOMEM;
 299 }
 300
 301 static void snmp6_free_dev(struct inet6_dev *idev)
 302 {
 303         snmp_mib_free((void __percpu **)idev->stats.icmpv6msg);
 304         snmp_mib_free((void __percpu **)idev->stats.icmpv6);
 305         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 306 }
 307
 308 /* Nobody refers to this device, we may destroy it. */
 309
 310 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 311 {
 312         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 313         kfree(idev);
 314 }
 315
 316 void in6_dev_finish_destroy(struct inet6_dev *idev)
 317 {
 318         struct net_device *dev = idev->dev;
 319
 320         WARN_ON(idev->addr_list != NULL);
 321         WARN_ON(idev->mc_list != NULL);
 322
 323 #ifdef NET_REFCNT_DEBUG
 324         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 325 #endif
 326         dev_put(dev);
 327         if (!idev->dead) {
 328                 printk("Freeing alive inet6 device %p\n", idev);
 329                 return;
 330         }
 331         snmp6_free_dev(idev);
 332         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 333 }
 334
 335 EXPORT_SYMBOL(in6_dev_finish_destroy);
 336
 337 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 338 {
 339         struct inet6_dev *ndev;
 340
 341         ASSERT_RTNL();
 342
 343         if (dev->mtu < IPV6_MIN_MTU)
 344                 return NULL;
 345
 346         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 347
 348         if (ndev == NULL)
 349                 return NULL;
 350
 351         rwlock_init(&ndev->lock);
 352         ndev->dev = dev;
 353         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 354         ndev->cnf.mtu6 = dev->mtu;
 355         ndev->cnf.sysctl = NULL;
 356         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 357         if (ndev->nd_parms == NULL) {
 358                 kfree(ndev);
 359                 return NULL;
 360         }
 361         if (ndev->cnf.forwarding)
 362                 dev_disable_lro(dev);
 363         /* We refer to the device */
 364         dev_hold(dev);
 365
 366         if (snmp6_alloc_dev(ndev) < 0) {
 367                 ADBG((KERN_WARNING
 368                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 369                         __func__, dev->name));
 370                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 371                 ndev->dead = 1;
 372                 in6_dev_finish_destroy(ndev);
 373                 return NULL;
 374         }
 375
 376         if (snmp6_register_dev(ndev) < 0) {
 377                 ADBG((KERN_WARNING
 378                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 379                         __func__, dev->name));
 380                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 381                 ndev->dead = 1;
 382                 in6_dev_finish_destroy(ndev);
 383                 return NULL;
 384         }
 385
 386         /* One reference from device.  We must do this before
 387          * we invoke __ipv6_regen_rndid().
 388          */
 389         in6_dev_hold(ndev);
 390
 391         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 392                 ndev->cnf.accept_dad = -1;
 393
 394 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 395         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 396                 printk(KERN_INFO
 397                        "%s: Disabled Multicast RS\n",
 398                        dev->name);
 399                 ndev->cnf.rtr_solicits = 0;
 400         }
 401 #endif
 402
 403 #ifdef CONFIG_IPV6_PRIVACY
 404         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 405         if ((dev->flags&IFF_LOOPBACK) ||
 406             dev->type == ARPHRD_TUNNEL ||
 407             dev->type == ARPHRD_TUNNEL6 ||
 408             dev->type == ARPHRD_SIT ||
 409             dev->type == ARPHRD_NONE) {
 410                 printk(KERN_INFO
 411                        "%s: Disabled Privacy Extensions\n",
 412                        dev->name);
 413                 ndev->cnf.use_tempaddr = -1;
 414         } else {
 415                 in6_dev_hold(ndev);
 416                 ipv6_regen_rndid((unsigned long) ndev);
 417         }
 418 #endif
 419
 420         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 421                 ndev->if_flags |= IF_READY;
 422
 423         ipv6_mc_init_dev(ndev);
 424         ndev->tstamp = jiffies;
 425         addrconf_sysctl_register(ndev);
 426         /* protected by rtnl_lock */
 427         rcu_assign_pointer(dev->ip6_ptr, ndev);
 428
 429         /* Join all-node multicast group */
 430         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 431
 432         return ndev;
 433 }
 434
 435 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 436 {
 437         struct inet6_dev *idev;
 438
 439         ASSERT_RTNL();
 440
 441         if ((idev = __in6_dev_get(dev)) == NULL) {
 442                 if ((idev = ipv6_add_dev(dev)) == NULL)
 443                         return NULL;
 444         }
 445
 446         if (dev->flags&IFF_UP)
 447                 ipv6_mc_up(idev);
 448         return idev;
 449 }
 450
 451 #ifdef CONFIG_SYSCTL
 452 static void dev_forward_change(struct inet6_dev *idev)
 453 {
 454         struct net_device *dev;
 455         struct inet6_ifaddr *ifa;
 456
 457         if (!idev)
 458                 return;
 459         dev = idev->dev;
 460         if (idev->cnf.forwarding)
 461                 dev_disable_lro(dev);
 462         if (dev && (dev->flags & IFF_MULTICAST)) {
 463                 if (idev->cnf.forwarding)
 464                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 465                 else
 466                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 467         }
 468         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
 469                 if (ifa->flags&IFA_F_TENTATIVE)
 470                         continue;
 471                 if (idev->cnf.forwarding)
 472                         addrconf_join_anycast(ifa);
 473                 else
 474                         addrconf_leave_anycast(ifa);
 475         }
 476 }
 477
 478
 479 static void addrconf_forward_change(struct net *net, __s32 newf)
 480 {
 481         struct net_device *dev;
 482         struct inet6_dev *idev;
 483
 484         rcu_read_lock();
 485         for_each_netdev_rcu(net, dev) {
 486                 idev = __in6_dev_get(dev);
 487                 if (idev) {
 488                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 489                         idev->cnf.forwarding = newf;
 490                         if (changed)
 491                                 dev_forward_change(idev);
 492                 }
 493         }
 494         rcu_read_unlock();
 495 }
 496
 497 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 498 {
 499         struct net *net;
 500
 501         net = (struct net *)table->extra2;
 502         if (p == &net->ipv6.devconf_dflt->forwarding)
 503                 return 0;
 504
 505         if (!rtnl_trylock()) {
 506                 /* Restore the original values before restarting */
 507                 *p = old;
 508                 return restart_syscall();
 509         }
 510
 511         if (p == &net->ipv6.devconf_all->forwarding) {
 512                 __s32 newf = net->ipv6.devconf_all->forwarding;
 513                 net->ipv6.devconf_dflt->forwarding = newf;
 514                 addrconf_forward_change(net, newf);
 515         } else if ((!*p) ^ (!old))
 516                 dev_forward_change((struct inet6_dev *)table->extra1);
 517         rtnl_unlock();
 518
 519         if (*p)
 520                 rt6_purge_dflt_routers(net);
 521         return 1;
 522 }
 523 #endif
 524
 525 /* Nobody refers to this ifaddr, destroy it */
 526
 527 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 528 {
 529         WARN_ON(ifp->if_next != NULL);
 530         WARN_ON(ifp->lst_next != NULL);
 531
 532 #ifdef NET_REFCNT_DEBUG
 533         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 534 #endif
 535
 536         in6_dev_put(ifp->idev);
 537
 538         if (del_timer(&ifp->timer))
 539                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
 540
 541         if (!ifp->dead) {
 542                 printk("Freeing alive inet6 address %p\n", ifp);
 543                 return;
 544         }
 545         dst_release(&ifp->rt->u.dst);
 546
 547         kfree(ifp);
 548 }
 549
 550 static void
 551 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 552 {
 553         struct inet6_ifaddr *ifa, **ifap;
 554         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 555
 556         /*
 557          * Each device address list is sorted in order of scope -
 558          * global before linklocal.
 559          */
 560         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
 561              ifap = &ifa->if_next) {
 562                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 563                         break;
 564         }
 565
 566         ifp->if_next = *ifap;
 567         *ifap = ifp;
 568 }
 569
 570 /*
 571  *      Hash function taken from net_alias.c
 572  */
 573 static u8 ipv6_addr_hash(const struct in6_addr *addr)
 574 {
 575         __u32 word;
 576
 577         /*
 578          * We perform the hash function over the last 64 bits of the address
 579          * This will include the IEEE address token on links that support it.
 580          */
 581
 582         word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
 583         word ^= (word >> 16);
 584         word ^= (word >> 8);
 585
 586         return ((word ^ (word >> 4)) & 0x0f);
 587 }
 588
 589 /* On success it returns ifp with increased reference count */
 590
 591 static struct inet6_ifaddr *
 592 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 593               int scope, u32 flags)
 594 {
 595         struct inet6_ifaddr *ifa = NULL;
 596         struct rt6_info *rt;
 597         int hash;
 598         int err = 0;
 599         int addr_type = ipv6_addr_type(addr);
 600
 601         if (addr_type == IPV6_ADDR_ANY ||
 602             addr_type & IPV6_ADDR_MULTICAST ||
 603             (!(idev->dev->flags & IFF_LOOPBACK) &&
 604              addr_type & IPV6_ADDR_LOOPBACK))
 605                 return ERR_PTR(-EADDRNOTAVAIL);
 606
 607         rcu_read_lock_bh();
 608         if (idev->dead) {
 609                 err = -ENODEV;                  /*XXX*/
 610                 goto out2;
 611         }
 612
 613         if (idev->cnf.disable_ipv6) {
 614                 err = -EACCES;
 615                 goto out2;
 616         }
 617
 618         write_lock(&addrconf_hash_lock);
 619
 620         /* Ignore adding duplicate addresses on an interface */
 621         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 622                 ADBG(("ipv6_add_addr: already assigned\n"));
 623                 err = -EEXIST;
 624                 goto out;
 625         }
 626
 627         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 628
 629         if (ifa == NULL) {
 630                 ADBG(("ipv6_add_addr: malloc failed\n"));
 631                 err = -ENOBUFS;
 632                 goto out;
 633         }
 634
 635         rt = addrconf_dst_alloc(idev, addr, 0);
 636         if (IS_ERR(rt)) {
 637                 err = PTR_ERR(rt);
 638                 goto out;
 639         }
 640
 641         ipv6_addr_copy(&ifa->addr, addr);
 642
 643         spin_lock_init(&ifa->lock);
 644         init_timer(&ifa->timer);
 645         ifa->timer.data = (unsigned long) ifa;
 646         ifa->scope = scope;
 647         ifa->prefix_len = pfxlen;
 648         ifa->flags = flags | IFA_F_TENTATIVE;
 649         ifa->cstamp = ifa->tstamp = jiffies;
 650
 651         ifa->rt = rt;
 652
 653         /*
 654          * part one of RFC 4429, section 3.3
 655          * We should not configure an address as
 656          * optimistic if we do not yet know the link
 657          * layer address of our nexhop router
 658          */
 659
 660         if (rt->rt6i_nexthop == NULL)
 661                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 662
 663         ifa->idev = idev;
 664         in6_dev_hold(idev);
 665         /* For caller */
 666         in6_ifa_hold(ifa);
 667
 668         /* Add to big hash table */
 669         hash = ipv6_addr_hash(addr);
 670
 671         ifa->lst_next = inet6_addr_lst[hash];
 672         inet6_addr_lst[hash] = ifa;
 673         in6_ifa_hold(ifa);
 674         write_unlock(&addrconf_hash_lock);
 675
 676         write_lock(&idev->lock);
 677         /* Add to inet6_dev unicast addr list. */
 678         ipv6_link_dev_addr(idev, ifa);
 679
 680 #ifdef CONFIG_IPV6_PRIVACY
 681         if (ifa->flags&IFA_F_TEMPORARY) {
 682                 ifa->tmp_next = idev->tempaddr_list;
 683                 idev->tempaddr_list = ifa;
 684                 in6_ifa_hold(ifa);
 685         }
 686 #endif
 687
 688         in6_ifa_hold(ifa);
 689         write_unlock(&idev->lock);
 690 out2:
 691         rcu_read_unlock_bh();
 692
 693         if (likely(err == 0))
 694                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 695         else {
 696                 kfree(ifa);
 697                 ifa = ERR_PTR(err);
 698         }
 699
 700         return ifa;
 701 out:
 702         write_unlock(&addrconf_hash_lock);
 703         goto out2;
 704 }
 705
 706 /* This function wants to get referenced ifp and releases it before return */
 707
 708 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 709 {
 710         struct inet6_ifaddr *ifa, **ifap;
 711         struct inet6_dev *idev = ifp->idev;
 712         int hash;
 713         int deleted = 0, onlink = 0;
 714         unsigned long expires = jiffies;
 715
 716         hash = ipv6_addr_hash(&ifp->addr);
 717
 718         ifp->dead = 1;
 719
 720         write_lock_bh(&addrconf_hash_lock);
 721         for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
 722              ifap = &ifa->lst_next) {
 723                 if (ifa == ifp) {
 724                         *ifap = ifa->lst_next;
 725                         __in6_ifa_put(ifp);
 726                         ifa->lst_next = NULL;
 727                         break;
 728                 }
 729         }
 730         write_unlock_bh(&addrconf_hash_lock);
 731
 732         write_lock_bh(&idev->lock);
 733 #ifdef CONFIG_IPV6_PRIVACY
 734         if (ifp->flags&IFA_F_TEMPORARY) {
 735                 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
 736                      ifap = &ifa->tmp_next) {
 737                         if (ifa == ifp) {
 738                                 *ifap = ifa->tmp_next;
 739                                 if (ifp->ifpub) {
 740                                         in6_ifa_put(ifp->ifpub);
 741                                         ifp->ifpub = NULL;
 742                                 }
 743                                 __in6_ifa_put(ifp);
 744                                 ifa->tmp_next = NULL;
 745                                 break;
 746                         }
 747                 }
 748         }
 749 #endif
 750
 751         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
 752                 if (ifa == ifp) {
 753                         *ifap = ifa->if_next;
 754                         __in6_ifa_put(ifp);
 755                         ifa->if_next = NULL;
 756                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 757                                 break;
 758                         deleted = 1;
 759                         continue;
 760                 } else if (ifp->flags & IFA_F_PERMANENT) {
 761                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 762                                               ifp->prefix_len)) {
 763                                 if (ifa->flags & IFA_F_PERMANENT) {
 764                                         onlink = 1;
 765                                         if (deleted)
 766                                                 break;
 767                                 } else {
 768                                         unsigned long lifetime;
 769
 770                                         if (!onlink)
 771                                                 onlink = -1;
 772
 773                                         spin_lock(&ifa->lock);
 774
 775                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 776                                         /*
 777                                          * Note: Because this address is
 778                                          * not permanent, lifetime <
 779                                          * LONG_MAX / HZ here.
 780                                          */
 781                                         if (time_before(expires,
 782                                                         ifa->tstamp + lifetime * HZ))
 783                                                 expires = ifa->tstamp + lifetime * HZ;
 784                                         spin_unlock(&ifa->lock);
 785                                 }
 786                         }
 787                 }
 788                 ifap = &ifa->if_next;
 789         }
 790         write_unlock_bh(&idev->lock);
 791
 792         addrconf_del_timer(ifp);
 793
 794         ipv6_ifa_notify(RTM_DELADDR, ifp);
 795
 796         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 797
 798         /*
 799          * Purge or update corresponding prefix
 800          *
 801          * 1) we don't purge prefix here if address was not permanent.
 802          *    prefix is managed by its own lifetime.
 803          * 2) if there're no addresses, delete prefix.
 804          * 3) if there're still other permanent address(es),
 805          *    corresponding prefix is still permanent.
 806          * 4) otherwise, update prefix lifetime to the
 807          *    longest valid lifetime among the corresponding
 808          *    addresses on the device.
 809          *    Note: subsequent RA will update lifetime.
 810          *
 811          * --yoshfuji
 812          */
 813         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 814                 struct in6_addr prefix;
 815                 struct rt6_info *rt;
 816                 struct net *net = dev_net(ifp->idev->dev);
 817                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 818                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 819
 820                 if (rt && addrconf_is_prefix_route(rt)) {
 821                         if (onlink == 0) {
 822                                 ip6_del_rt(rt);
 823                                 rt = NULL;
 824                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 825                                 rt->rt6i_expires = expires;
 826                                 rt->rt6i_flags |= RTF_EXPIRES;
 827                         }
 828                 }
 829                 dst_release(&rt->u.dst);
 830         }
 831
 832         in6_ifa_put(ifp);
 833 }
 834
 835 #ifdef CONFIG_IPV6_PRIVACY
 836 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 837 {
 838         struct inet6_dev *idev = ifp->idev;
 839         struct in6_addr addr, *tmpaddr;
 840         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
 841         unsigned long regen_advance;
 842         int tmp_plen;
 843         int ret = 0;
 844         int max_addresses;
 845         u32 addr_flags;
 846
 847         write_lock(&idev->lock);
 848         if (ift) {
 849                 spin_lock_bh(&ift->lock);
 850                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 851                 spin_unlock_bh(&ift->lock);
 852                 tmpaddr = &addr;
 853         } else {
 854                 tmpaddr = NULL;
 855         }
 856 retry:
 857         in6_dev_hold(idev);
 858         if (idev->cnf.use_tempaddr <= 0) {
 859                 write_unlock(&idev->lock);
 860                 printk(KERN_INFO
 861                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 862                 in6_dev_put(idev);
 863                 ret = -1;
 864                 goto out;
 865         }
 866         spin_lock_bh(&ifp->lock);
 867         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 868                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 869                 spin_unlock_bh(&ifp->lock);
 870                 write_unlock(&idev->lock);
 871                 printk(KERN_WARNING
 872                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 873                 in6_dev_put(idev);
 874                 ret = -1;
 875                 goto out;
 876         }
 877         in6_ifa_hold(ifp);
 878         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 879         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 880                 spin_unlock_bh(&ifp->lock);
 881                 write_unlock(&idev->lock);
 882                 printk(KERN_WARNING
 883                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 884                 in6_ifa_put(ifp);
 885                 in6_dev_put(idev);
 886                 ret = -1;
 887                 goto out;
 888         }
 889         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 890         tmp_valid_lft = min_t(__u32,
 891                               ifp->valid_lft,
 892                               idev->cnf.temp_valid_lft);
 893         tmp_prefered_lft = min_t(__u32,
 894                                  ifp->prefered_lft,
 895                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
 896         tmp_plen = ifp->prefix_len;
 897         max_addresses = idev->cnf.max_addresses;
 898         tmp_cstamp = ifp->cstamp;
 899         tmp_tstamp = ifp->tstamp;
 900         spin_unlock_bh(&ifp->lock);
 901
 902         regen_advance = idev->cnf.regen_max_retry *
 903                         idev->cnf.dad_transmits *
 904                         idev->nd_parms->retrans_time / HZ;
 905         write_unlock(&idev->lock);
 906
 907         /* A temporary address is created only if this calculated Preferred
 908          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 909          * an implementation must not create a temporary address with a zero
 910          * Preferred Lifetime.
 911          */
 912         if (tmp_prefered_lft <= regen_advance) {
 913                 in6_ifa_put(ifp);
 914                 in6_dev_put(idev);
 915                 ret = -1;
 916                 goto out;
 917         }
 918
 919         addr_flags = IFA_F_TEMPORARY;
 920         /* set in addrconf_prefix_rcv() */
 921         if (ifp->flags & IFA_F_OPTIMISTIC)
 922                 addr_flags |= IFA_F_OPTIMISTIC;
 923
 924         ift = !max_addresses ||
 925               ipv6_count_addresses(idev) < max_addresses ?
 926                 ipv6_add_addr(idev, &addr, tmp_plen,
 927                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 928                               addr_flags) : NULL;
 929         if (!ift || IS_ERR(ift)) {
 930                 in6_ifa_put(ifp);
 931                 in6_dev_put(idev);
 932                 printk(KERN_INFO
 933                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 934                 tmpaddr = &addr;
 935                 write_lock(&idev->lock);
 936                 goto retry;
 937         }
 938
 939         spin_lock_bh(&ift->lock);
 940         ift->ifpub = ifp;
 941         ift->valid_lft = tmp_valid_lft;
 942         ift->prefered_lft = tmp_prefered_lft;
 943         ift->cstamp = tmp_cstamp;
 944         ift->tstamp = tmp_tstamp;
 945         spin_unlock_bh(&ift->lock);
 946
 947         addrconf_dad_start(ift, 0);
 948         in6_ifa_put(ift);
 949         in6_dev_put(idev);
 950 out:
 951         return ret;
 952 }
 953 #endif
 954
 955 /*
 956  *      Choose an appropriate source address (RFC3484)
 957  */
 958 enum {
 959         IPV6_SADDR_RULE_INIT = 0,
 960         IPV6_SADDR_RULE_LOCAL,
 961         IPV6_SADDR_RULE_SCOPE,
 962         IPV6_SADDR_RULE_PREFERRED,
 963 #ifdef CONFIG_IPV6_MIP6
 964         IPV6_SADDR_RULE_HOA,
 965 #endif
 966         IPV6_SADDR_RULE_OIF,
 967         IPV6_SADDR_RULE_LABEL,
 968 #ifdef CONFIG_IPV6_PRIVACY
 969         IPV6_SADDR_RULE_PRIVACY,
 970 #endif
 971         IPV6_SADDR_RULE_ORCHID,
 972         IPV6_SADDR_RULE_PREFIX,
 973         IPV6_SADDR_RULE_MAX
 974 };
 975
 976 struct ipv6_saddr_score {
 977         int                     rule;
 978         int                     addr_type;
 979         struct inet6_ifaddr     *ifa;
 980         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 981         int                     scopedist;
 982         int                     matchlen;
 983 };
 984
 985 struct ipv6_saddr_dst {
 986         const struct in6_addr *addr;
 987         int ifindex;
 988         int scope;
 989         int label;
 990         unsigned int prefs;
 991 };
 992
 993 static inline int ipv6_saddr_preferred(int type)
 994 {
 995         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 996                 return 1;
 997         return 0;
 998 }
 999
1000 static int ipv6_get_saddr_eval(struct net *net,
1001                                struct ipv6_saddr_score *score,
1002                                struct ipv6_saddr_dst *dst,
1003                                int i)
1004 {
1005         int ret;
1006
1007         if (i <= score->rule) {
1008                 switch (i) {
1009                 case IPV6_SADDR_RULE_SCOPE:
1010                         ret = score->scopedist;
1011                         break;
1012                 case IPV6_SADDR_RULE_PREFIX:
1013                         ret = score->matchlen;
1014                         break;
1015                 default:
1016                         ret = !!test_bit(i, score->scorebits);
1017                 }
1018                 goto out;
1019         }
1020
1021         switch (i) {
1022         case IPV6_SADDR_RULE_INIT:
1023                 /* Rule 0: remember if hiscore is not ready yet */
1024                 ret = !!score->ifa;
1025                 break;
1026         case IPV6_SADDR_RULE_LOCAL:
1027                 /* Rule 1: Prefer same address */
1028                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1029                 break;
1030         case IPV6_SADDR_RULE_SCOPE:
1031                 /* Rule 2: Prefer appropriate scope
1032                  *
1033                  *      ret
1034                  *       ^
1035                  *    -1 |  d 15
1036                  *    ---+--+-+---> scope
1037                  *       |
1038                  *       |             d is scope of the destination.
1039                  *  B-d  |  \
1040                  *       |   \      <- smaller scope is better if
1041                  *  B-15 |    \        if scope is enough for destinaion.
1042                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1043                  * d-C-1 | /
1044                  *       |/         <- greater is better
1045                  *   -C  /             if scope is not enough for destination.
1046                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1047                  *
1048                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1049                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1050                  * Assume B = 0 and we get C > 29.
1051                  */
1052                 ret = __ipv6_addr_src_scope(score->addr_type);
1053                 if (ret >= dst->scope)
1054                         ret = -ret;
1055                 else
1056                         ret -= 128;     /* 30 is enough */
1057                 score->scopedist = ret;
1058                 break;
1059         case IPV6_SADDR_RULE_PREFERRED:
1060                 /* Rule 3: Avoid deprecated and optimistic addresses */
1061                 ret = ipv6_saddr_preferred(score->addr_type) ||
1062                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1063                 break;
1064 #ifdef CONFIG_IPV6_MIP6
1065         case IPV6_SADDR_RULE_HOA:
1066             {
1067                 /* Rule 4: Prefer home address */
1068                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1069                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1070                 break;
1071             }
1072 #endif
1073         case IPV6_SADDR_RULE_OIF:
1074                 /* Rule 5: Prefer outgoing interface */
1075                 ret = (!dst->ifindex ||
1076                        dst->ifindex == score->ifa->idev->dev->ifindex);
1077                 break;
1078         case IPV6_SADDR_RULE_LABEL:
1079                 /* Rule 6: Prefer matching label */
1080                 ret = ipv6_addr_label(net,
1081                                       &score->ifa->addr, score->addr_type,
1082                                       score->ifa->idev->dev->ifindex) == dst->label;
1083                 break;
1084 #ifdef CONFIG_IPV6_PRIVACY
1085         case IPV6_SADDR_RULE_PRIVACY:
1086             {
1087                 /* Rule 7: Prefer public address
1088                  * Note: prefer temprary address if use_tempaddr >= 2
1089                  */
1090                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1091                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1092                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1093                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1094                 break;
1095             }
1096 #endif
1097         case IPV6_SADDR_RULE_ORCHID:
1098                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1099                  *          non-ORCHID vs non-ORCHID
1100                  */
1101                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1102                         ipv6_addr_orchid(dst->addr));
1103                 break;
1104         case IPV6_SADDR_RULE_PREFIX:
1105                 /* Rule 8: Use longest matching prefix */
1106                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1107                                                        dst->addr);
1108                 break;
1109         default:
1110                 ret = 0;
1111         }
1112
1113         if (ret)
1114                 __set_bit(i, score->scorebits);
1115         score->rule = i;
1116 out:
1117         return ret;
1118 }
1119
1120 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1121                        const struct in6_addr *daddr, unsigned int prefs,
1122                        struct in6_addr *saddr)
1123 {
1124         struct ipv6_saddr_score scores[2],
1125                                 *score = &scores[0], *hiscore = &scores[1];
1126         struct ipv6_saddr_dst dst;
1127         struct net_device *dev;
1128         int dst_type;
1129
1130         dst_type = __ipv6_addr_type(daddr);
1131         dst.addr = daddr;
1132         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1133         dst.scope = __ipv6_addr_src_scope(dst_type);
1134         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1135         dst.prefs = prefs;
1136
1137         hiscore->rule = -1;
1138         hiscore->ifa = NULL;
1139
1140         rcu_read_lock();
1141
1142         for_each_netdev_rcu(net, dev) {
1143                 struct inet6_dev *idev;
1144
1145                 /* Candidate Source Address (section 4)
1146                  *  - multicast and link-local destination address,
1147                  *    the set of candidate source address MUST only
1148                  *    include addresses assigned to interfaces
1149                  *    belonging to the same link as the outgoing
1150                  *    interface.
1151                  * (- For site-local destination addresses, the
1152                  *    set of candidate source addresses MUST only
1153                  *    include addresses assigned to interfaces
1154                  *    belonging to the same site as the outgoing
1155                  *    interface.)
1156                  */
1157                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1158                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1159                     dst.ifindex && dev->ifindex != dst.ifindex)
1160                         continue;
1161
1162                 idev = __in6_dev_get(dev);
1163                 if (!idev)
1164                         continue;
1165
1166                 read_lock_bh(&idev->lock);
1167                 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1168                         int i;
1169
1170                         /*
1171                          * - Tentative Address (RFC2462 section 5.4)
1172                          *  - A tentative address is not considered
1173                          *    "assigned to an interface" in the traditional
1174                          *    sense, unless it is also flagged as optimistic.
1175                          * - Candidate Source Address (section 4)
1176                          *  - In any case, anycast addresses, multicast
1177                          *    addresses, and the unspecified address MUST
1178                          *    NOT be included in a candidate set.
1179                          */
1180                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1181                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1182                                 continue;
1183
1184                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1185
1186                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1187                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1188                                 LIMIT_NETDEBUG(KERN_DEBUG
1189                                                "ADDRCONF: unspecified / multicast address "
1190                                                "assigned as unicast address on %s",
1191                                                dev->name);
1192                                 continue;
1193                         }
1194
1195                         score->rule = -1;
1196                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1197
1198                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1199                                 int minihiscore, miniscore;
1200
1201                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1202                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1203
1204                                 if (minihiscore > miniscore) {
1205                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1206                                             score->scopedist > 0) {
1207                                                 /*
1208                                                  * special case:
1209                                                  * each remaining entry
1210                                                  * has too small (not enough)
1211                                                  * scope, because ifa entries
1212                                                  * are sorted by their scope
1213                                                  * values.
1214                                                  */
1215                                                 goto try_nextdev;
1216                                         }
1217                                         break;
1218                                 } else if (minihiscore < miniscore) {
1219                                         if (hiscore->ifa)
1220                                                 in6_ifa_put(hiscore->ifa);
1221
1222                                         in6_ifa_hold(score->ifa);
1223
1224                                         swap(hiscore, score);
1225
1226                                         /* restore our iterator */
1227                                         score->ifa = hiscore->ifa;
1228
1229                                         break;
1230                                 }
1231                         }
1232                 }
1233 try_nextdev:
1234                 read_unlock_bh(&idev->lock);
1235         }
1236         rcu_read_unlock();
1237
1238         if (!hiscore->ifa)
1239                 return -EADDRNOTAVAIL;
1240
1241         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1242         in6_ifa_put(hiscore->ifa);
1243         return 0;
1244 }
1245
1246 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1247
1248 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1249                     unsigned char banned_flags)
1250 {
1251         struct inet6_dev *idev;
1252         int err = -EADDRNOTAVAIL;
1253
1254         rcu_read_lock();
1255         if ((idev = __in6_dev_get(dev)) != NULL) {
1256                 struct inet6_ifaddr *ifp;
1257
1258                 read_lock_bh(&idev->lock);
1259                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1260                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1261                                 ipv6_addr_copy(addr, &ifp->addr);
1262                                 err = 0;
1263                                 break;
1264                         }
1265                 }
1266                 read_unlock_bh(&idev->lock);
1267         }
1268         rcu_read_unlock();
1269         return err;
1270 }
1271
1272 static int ipv6_count_addresses(struct inet6_dev *idev)
1273 {
1274         int cnt = 0;
1275         struct inet6_ifaddr *ifp;
1276
1277         read_lock_bh(&idev->lock);
1278         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1279                 cnt++;
1280         read_unlock_bh(&idev->lock);
1281         return cnt;
1282 }
1283
1284 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1285                   struct net_device *dev, int strict)
1286 {
1287         struct inet6_ifaddr * ifp;
1288         u8 hash = ipv6_addr_hash(addr);
1289
1290         read_lock_bh(&addrconf_hash_lock);
1291         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1292                 if (!net_eq(dev_net(ifp->idev->dev), net))
1293                         continue;
1294                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1295                     !(ifp->flags&IFA_F_TENTATIVE)) {
1296                         if (dev == NULL || ifp->idev->dev == dev ||
1297                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1298                                 break;
1299                 }
1300         }
1301         read_unlock_bh(&addrconf_hash_lock);
1302         return ifp != NULL;
1303 }
1304 EXPORT_SYMBOL(ipv6_chk_addr);
1305
1306 static
1307 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1308                        struct net_device *dev)
1309 {
1310         struct inet6_ifaddr * ifp;
1311         u8 hash = ipv6_addr_hash(addr);
1312
1313         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1314                 if (!net_eq(dev_net(ifp->idev->dev), net))
1315                         continue;
1316                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1317                         if (dev == NULL || ifp->idev->dev == dev)
1318                                 break;
1319                 }
1320         }
1321         return ifp != NULL;
1322 }
1323
1324 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1325 {
1326         struct inet6_dev *idev;
1327         struct inet6_ifaddr *ifa;
1328         int     onlink;
1329
1330         onlink = 0;
1331         rcu_read_lock();
1332         idev = __in6_dev_get(dev);
1333         if (idev) {
1334                 read_lock_bh(&idev->lock);
1335                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1336                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1337                                                    ifa->prefix_len);
1338                         if (onlink)
1339                                 break;
1340                 }
1341                 read_unlock_bh(&idev->lock);
1342         }
1343         rcu_read_unlock();
1344         return onlink;
1345 }
1346
1347 EXPORT_SYMBOL(ipv6_chk_prefix);
1348
1349 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1350                                      struct net_device *dev, int strict)
1351 {
1352         struct inet6_ifaddr * ifp;
1353         u8 hash = ipv6_addr_hash(addr);
1354
1355         read_lock_bh(&addrconf_hash_lock);
1356         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1357                 if (!net_eq(dev_net(ifp->idev->dev), net))
1358                         continue;
1359                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1360                         if (dev == NULL || ifp->idev->dev == dev ||
1361                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1362                                 in6_ifa_hold(ifp);
1363                                 break;
1364                         }
1365                 }
1366         }
1367         read_unlock_bh(&addrconf_hash_lock);
1368
1369         return ifp;
1370 }
1371
1372 /* Gets referenced address, destroys ifaddr */
1373
1374 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1375 {
1376         if (ifp->flags&IFA_F_PERMANENT) {
1377                 spin_lock_bh(&ifp->lock);
1378                 addrconf_del_timer(ifp);
1379                 ifp->flags |= IFA_F_TENTATIVE;
1380                 if (dad_failed)
1381                         ifp->flags |= IFA_F_DADFAILED;
1382                 spin_unlock_bh(&ifp->lock);
1383                 in6_ifa_put(ifp);
1384 #ifdef CONFIG_IPV6_PRIVACY
1385         } else if (ifp->flags&IFA_F_TEMPORARY) {
1386                 struct inet6_ifaddr *ifpub;
1387                 spin_lock_bh(&ifp->lock);
1388                 ifpub = ifp->ifpub;
1389                 if (ifpub) {
1390                         in6_ifa_hold(ifpub);
1391                         spin_unlock_bh(&ifp->lock);
1392                         ipv6_create_tempaddr(ifpub, ifp);
1393                         in6_ifa_put(ifpub);
1394                 } else {
1395                         spin_unlock_bh(&ifp->lock);
1396                 }
1397                 ipv6_del_addr(ifp);
1398 #endif
1399         } else
1400                 ipv6_del_addr(ifp);
1401 }
1402
1403 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1404 {
1405         struct inet6_dev *idev = ifp->idev;
1406
1407         if (net_ratelimit())
1408                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1409                         ifp->idev->dev->name, &ifp->addr);
1410
1411         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1412                 struct in6_addr addr;
1413
1414                 addr.s6_addr32[0] = htonl(0xfe800000);
1415                 addr.s6_addr32[1] = 0;
1416
1417                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1418                     ipv6_addr_equal(&ifp->addr, &addr)) {
1419                         /* DAD failed for link-local based on MAC address */
1420                         idev->cnf.disable_ipv6 = 1;
1421
1422                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1423                                 ifp->idev->dev->name);
1424                 }
1425         }
1426
1427         addrconf_dad_stop(ifp, 1);
1428 }
1429
1430 /* Join to solicited addr multicast group. */
1431
1432 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1433 {
1434         struct in6_addr maddr;
1435
1436         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1437                 return;
1438
1439         addrconf_addr_solict_mult(addr, &maddr);
1440         ipv6_dev_mc_inc(dev, &maddr);
1441 }
1442
1443 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1444 {
1445         struct in6_addr maddr;
1446
1447         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1448                 return;
1449
1450         addrconf_addr_solict_mult(addr, &maddr);
1451         __ipv6_dev_mc_dec(idev, &maddr);
1452 }
1453
1454 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1455 {
1456         struct in6_addr addr;
1457         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1458         if (ipv6_addr_any(&addr))
1459                 return;
1460         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1461 }
1462
1463 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1464 {
1465         struct in6_addr addr;
1466         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1467         if (ipv6_addr_any(&addr))
1468                 return;
1469         __ipv6_dev_ac_dec(ifp->idev, &addr);
1470 }
1471
1472 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1473 {
1474         if (dev->addr_len != ETH_ALEN)
1475                 return -1;
1476         memcpy(eui, dev->dev_addr, 3);
1477         memcpy(eui + 5, dev->dev_addr + 3, 3);
1478
1479         /*
1480          * The zSeries OSA network cards can be shared among various
1481          * OS instances, but the OSA cards have only one MAC address.
1482          * This leads to duplicate address conflicts in conjunction
1483          * with IPv6 if more than one instance uses the same card.
1484          *
1485          * The driver for these cards can deliver a unique 16-bit
1486          * identifier for each instance sharing the same card.  It is
1487          * placed instead of 0xFFFE in the interface identifier.  The
1488          * "u" bit of the interface identifier is not inverted in this
1489          * case.  Hence the resulting interface identifier has local
1490          * scope according to RFC2373.
1491          */
1492         if (dev->dev_id) {
1493                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1494                 eui[4] = dev->dev_id & 0xFF;
1495         } else {
1496                 eui[3] = 0xFF;
1497                 eui[4] = 0xFE;
1498                 eui[0] ^= 2;
1499         }
1500         return 0;
1501 }
1502
1503 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1504 {
1505         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1506         if (dev->addr_len != ARCNET_ALEN)
1507                 return -1;
1508         memset(eui, 0, 7);
1509         eui[7] = *(u8*)dev->dev_addr;
1510         return 0;
1511 }
1512
1513 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1514 {
1515         if (dev->addr_len != INFINIBAND_ALEN)
1516                 return -1;
1517         memcpy(eui, dev->dev_addr + 12, 8);
1518         eui[0] |= 2;
1519         return 0;
1520 }
1521
1522 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1523 {
1524         if (addr == 0)
1525                 return -1;
1526         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1527                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1528                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1529                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1530                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1531                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1532         eui[1] = 0;
1533         eui[2] = 0x5E;
1534         eui[3] = 0xFE;
1535         memcpy(eui + 4, &addr, 4);
1536         return 0;
1537 }
1538 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1539
1540 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1541 {
1542         if (dev->priv_flags & IFF_ISATAP)
1543                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1544         return -1;
1545 }
1546
1547 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1548 {
1549         switch (dev->type) {
1550         case ARPHRD_ETHER:
1551         case ARPHRD_FDDI:
1552         case ARPHRD_IEEE802_TR:
1553                 return addrconf_ifid_eui48(eui, dev);
1554         case ARPHRD_ARCNET:
1555                 return addrconf_ifid_arcnet(eui, dev);
1556         case ARPHRD_INFINIBAND:
1557                 return addrconf_ifid_infiniband(eui, dev);
1558         case ARPHRD_SIT:
1559                 return addrconf_ifid_sit(eui, dev);
1560         }
1561         return -1;
1562 }
1563
1564 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1565 {
1566         int err = -1;
1567         struct inet6_ifaddr *ifp;
1568
1569         read_lock_bh(&idev->lock);
1570         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1571                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1572                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1573                         err = 0;
1574                         break;
1575                 }
1576         }
1577         read_unlock_bh(&idev->lock);
1578         return err;
1579 }
1580
1581 #ifdef CONFIG_IPV6_PRIVACY
1582 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1583 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1584 {
1585 regen:
1586         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1587         idev->rndid[0] &= ~0x02;
1588
1589         /*
1590          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1591          * check if generated address is not inappropriate
1592          *
1593          *  - Reserved subnet anycast (RFC 2526)
1594          *      11111101 11....11 1xxxxxxx
1595          *  - ISATAP (RFC4214) 6.1
1596          *      00-00-5E-FE-xx-xx-xx-xx
1597          *  - value 0
1598          *  - XXX: already assigned to an address on the device
1599          */
1600         if (idev->rndid[0] == 0xfd &&
1601             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1602             (idev->rndid[7]&0x80))
1603                 goto regen;
1604         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1605                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1606                         goto regen;
1607                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1608                         goto regen;
1609         }
1610
1611         return 0;
1612 }
1613
1614 static void ipv6_regen_rndid(unsigned long data)
1615 {
1616         struct inet6_dev *idev = (struct inet6_dev *) data;
1617         unsigned long expires;
1618
1619         rcu_read_lock_bh();
1620         write_lock_bh(&idev->lock);
1621
1622         if (idev->dead)
1623                 goto out;
1624
1625         if (__ipv6_regen_rndid(idev) < 0)
1626                 goto out;
1627
1628         expires = jiffies +
1629                 idev->cnf.temp_prefered_lft * HZ -
1630                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1631         if (time_before(expires, jiffies)) {
1632                 printk(KERN_WARNING
1633                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1634                         idev->dev->name);
1635                 goto out;
1636         }
1637
1638         if (!mod_timer(&idev->regen_timer, expires))
1639                 in6_dev_hold(idev);
1640
1641 out:
1642         write_unlock_bh(&idev->lock);
1643         rcu_read_unlock_bh();
1644         in6_dev_put(idev);
1645 }
1646
1647 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1648         int ret = 0;
1649
1650         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1651                 ret = __ipv6_regen_rndid(idev);
1652         return ret;
1653 }
1654 #endif
1655
1656 /*
1657  *      Add prefix route.
1658  */
1659
1660 static void
1661 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1662                       unsigned long expires, u32 flags)
1663 {
1664         struct fib6_config cfg = {
1665                 .fc_table = RT6_TABLE_PREFIX,
1666                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1667                 .fc_ifindex = dev->ifindex,
1668                 .fc_expires = expires,
1669                 .fc_dst_len = plen,
1670                 .fc_flags = RTF_UP | flags,
1671                 .fc_nlinfo.nl_net = dev_net(dev),
1672                 .fc_protocol = RTPROT_KERNEL,
1673         };
1674
1675         ipv6_addr_copy(&cfg.fc_dst, pfx);
1676
1677         /* Prevent useless cloning on PtP SIT.
1678            This thing is done here expecting that the whole
1679            class of non-broadcast devices need not cloning.
1680          */
1681 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1682         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1683                 cfg.fc_flags |= RTF_NONEXTHOP;
1684 #endif
1685
1686         ip6_route_add(&cfg);
1687 }
1688
1689 /* Create "default" multicast route to the interface */
1690
1691 static void addrconf_add_mroute(struct net_device *dev)
1692 {
1693         struct fib6_config cfg = {
1694                 .fc_table = RT6_TABLE_LOCAL,
1695                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1696                 .fc_ifindex = dev->ifindex,
1697                 .fc_dst_len = 8,
1698                 .fc_flags = RTF_UP,
1699                 .fc_nlinfo.nl_net = dev_net(dev),
1700         };
1701
1702         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1703
1704         ip6_route_add(&cfg);
1705 }
1706
1707 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1708 static void sit_route_add(struct net_device *dev)
1709 {
1710         struct fib6_config cfg = {
1711                 .fc_table = RT6_TABLE_MAIN,
1712                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1713                 .fc_ifindex = dev->ifindex,
1714                 .fc_dst_len = 96,
1715                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1716                 .fc_nlinfo.nl_net = dev_net(dev),
1717         };
1718
1719         /* prefix length - 96 bits "::d.d.d.d" */
1720         ip6_route_add(&cfg);
1721 }
1722 #endif
1723
1724 static void addrconf_add_lroute(struct net_device *dev)
1725 {
1726         struct in6_addr addr;
1727
1728         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1729         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1730 }
1731
1732 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1733 {
1734         struct inet6_dev *idev;
1735
1736         ASSERT_RTNL();
1737
1738         if ((idev = ipv6_find_idev(dev)) == NULL)
1739                 return NULL;
1740
1741         /* Add default multicast route */
1742         addrconf_add_mroute(dev);
1743
1744         /* Add link local route */
1745         addrconf_add_lroute(dev);
1746         return idev;
1747 }
1748
1749 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1750 {
1751         struct prefix_info *pinfo;
1752         __u32 valid_lft;
1753         __u32 prefered_lft;
1754         int addr_type;
1755         struct inet6_dev *in6_dev;
1756         struct net *net = dev_net(dev);
1757
1758         pinfo = (struct prefix_info *) opt;
1759
1760         if (len < sizeof(struct prefix_info)) {
1761                 ADBG(("addrconf: prefix option too short\n"));
1762                 return;
1763         }
1764
1765         /*
1766          *      Validation checks ([ADDRCONF], page 19)
1767          */
1768
1769         addr_type = ipv6_addr_type(&pinfo->prefix);
1770
1771         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1772                 return;
1773
1774         valid_lft = ntohl(pinfo->valid);
1775         prefered_lft = ntohl(pinfo->prefered);
1776
1777         if (prefered_lft > valid_lft) {
1778                 if (net_ratelimit())
1779                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1780                 return;
1781         }
1782
1783         in6_dev = in6_dev_get(dev);
1784
1785         if (in6_dev == NULL) {
1786                 if (net_ratelimit())
1787                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1788                 return;
1789         }
1790
1791         /*
1792          *      Two things going on here:
1793          *      1) Add routes for on-link prefixes
1794          *      2) Configure prefixes with the auto flag set
1795          */
1796
1797         if (pinfo->onlink) {
1798                 struct rt6_info *rt;
1799                 unsigned long rt_expires;
1800
1801                 /* Avoid arithmetic overflow. Really, we could
1802                  * save rt_expires in seconds, likely valid_lft,
1803                  * but it would require division in fib gc, that it
1804                  * not good.
1805                  */
1806                 if (HZ > USER_HZ)
1807                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1808                 else
1809                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1810
1811                 if (addrconf_finite_timeout(rt_expires))
1812                         rt_expires *= HZ;
1813
1814                 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1815                                 dev->ifindex, 1);
1816
1817                 if (rt && addrconf_is_prefix_route(rt)) {
1818                         /* Autoconf prefix route */
1819                         if (valid_lft == 0) {
1820                                 ip6_del_rt(rt);
1821                                 rt = NULL;
1822                         } else if (addrconf_finite_timeout(rt_expires)) {
1823                                 /* not infinity */
1824                                 rt->rt6i_expires = jiffies + rt_expires;
1825                                 rt->rt6i_flags |= RTF_EXPIRES;
1826                         } else {
1827                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1828                                 rt->rt6i_expires = 0;
1829                         }
1830                 } else if (valid_lft) {
1831                         clock_t expires = 0;
1832                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1833                         if (addrconf_finite_timeout(rt_expires)) {
1834                                 /* not infinity */
1835                                 flags |= RTF_EXPIRES;
1836                                 expires = jiffies_to_clock_t(rt_expires);
1837                         }
1838                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1839                                               dev, expires, flags);
1840                 }
1841                 if (rt)
1842                         dst_release(&rt->u.dst);
1843         }
1844
1845         /* Try to figure out our local address for this prefix */
1846
1847         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1848                 struct inet6_ifaddr * ifp;
1849                 struct in6_addr addr;
1850                 int create = 0, update_lft = 0;
1851
1852                 if (pinfo->prefix_len == 64) {
1853                         memcpy(&addr, &pinfo->prefix, 8);
1854                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1855                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1856                                 in6_dev_put(in6_dev);
1857                                 return;
1858                         }
1859                         goto ok;
1860                 }
1861                 if (net_ratelimit())
1862                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1863                                pinfo->prefix_len);
1864                 in6_dev_put(in6_dev);
1865                 return;
1866
1867 ok:
1868
1869                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1870
1871                 if (ifp == NULL && valid_lft) {
1872                         int max_addresses = in6_dev->cnf.max_addresses;
1873                         u32 addr_flags = 0;
1874
1875 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1876                         if (in6_dev->cnf.optimistic_dad &&
1877                             !net->ipv6.devconf_all->forwarding)
1878                                 addr_flags = IFA_F_OPTIMISTIC;
1879 #endif
1880
1881                         /* Do not allow to create too much of autoconfigured
1882                          * addresses; this would be too easy way to crash kernel.
1883                          */
1884                         if (!max_addresses ||
1885                             ipv6_count_addresses(in6_dev) < max_addresses)
1886                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1887                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1888                                                     addr_flags);
1889
1890                         if (!ifp || IS_ERR(ifp)) {
1891                                 in6_dev_put(in6_dev);
1892                                 return;
1893                         }
1894
1895                         update_lft = create = 1;
1896                         ifp->cstamp = jiffies;
1897                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1898                 }
1899
1900                 if (ifp) {
1901                         int flags;
1902                         unsigned long now;
1903 #ifdef CONFIG_IPV6_PRIVACY
1904                         struct inet6_ifaddr *ift;
1905 #endif
1906                         u32 stored_lft;
1907
1908                         /* update lifetime (RFC2462 5.5.3 e) */
1909                         spin_lock(&ifp->lock);
1910                         now = jiffies;
1911                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1912                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1913                         else
1914                                 stored_lft = 0;
1915                         if (!update_lft && stored_lft) {
1916                                 if (valid_lft > MIN_VALID_LIFETIME ||
1917                                     valid_lft > stored_lft)
1918                                         update_lft = 1;
1919                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1920                                         /* valid_lft <= stored_lft is always true */
1921                                         /*
1922                                          * RFC 4862 Section 5.5.3e:
1923                                          * "Note that the preferred lifetime of
1924                                          *  the corresponding address is always
1925                                          *  reset to the Preferred Lifetime in
1926                                          *  the received Prefix Information
1927                                          *  option, regardless of whether the
1928                                          *  valid lifetime is also reset or
1929                                          *  ignored."
1930                                          *
1931                                          *  So if the preferred lifetime in
1932                                          *  this advertisement is different
1933                                          *  than what we have stored, but the
1934                                          *  valid lifetime is invalid, just
1935                                          *  reset prefered_lft.
1936                                          *
1937                                          *  We must set the valid lifetime
1938                                          *  to the stored lifetime since we'll
1939                                          *  be updating the timestamp below,
1940                                          *  else we'll set it back to the
1941                                          *  minumum.
1942                                          */
1943                                         if (prefered_lft != ifp->prefered_lft) {
1944                                                 valid_lft = stored_lft;
1945                                                 update_lft = 1;
1946                                         }
1947                                 } else {
1948                                         valid_lft = MIN_VALID_LIFETIME;
1949                                         if (valid_lft < prefered_lft)
1950                                                 prefered_lft = valid_lft;
1951                                         update_lft = 1;
1952                                 }
1953                         }
1954
1955                         if (update_lft) {
1956                                 ifp->valid_lft = valid_lft;
1957                                 ifp->prefered_lft = prefered_lft;
1958                                 ifp->tstamp = now;
1959                                 flags = ifp->flags;
1960                                 ifp->flags &= ~IFA_F_DEPRECATED;
1961                                 spin_unlock(&ifp->lock);
1962
1963                                 if (!(flags&IFA_F_TENTATIVE))
1964                                         ipv6_ifa_notify(0, ifp);
1965                         } else
1966                                 spin_unlock(&ifp->lock);
1967
1968 #ifdef CONFIG_IPV6_PRIVACY
1969                         read_lock_bh(&in6_dev->lock);
1970                         /* update all temporary addresses in the list */
1971                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1972                                 /*
1973                                  * When adjusting the lifetimes of an existing
1974                                  * temporary address, only lower the lifetimes.
1975                                  * Implementations must not increase the
1976                                  * lifetimes of an existing temporary address
1977                                  * when processing a Prefix Information Option.
1978                                  */
1979                                 if (ifp != ift->ifpub)
1980                                         continue;
1981
1982                                 spin_lock(&ift->lock);
1983                                 flags = ift->flags;
1984                                 if (ift->valid_lft > valid_lft &&
1985                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1986                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1987                                 if (ift->prefered_lft > prefered_lft &&
1988                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1989                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1990                                 spin_unlock(&ift->lock);
1991                                 if (!(flags&IFA_F_TENTATIVE))
1992                                         ipv6_ifa_notify(0, ift);
1993                         }
1994
1995                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1996                                 /*
1997                                  * When a new public address is created as described in [ADDRCONF],
1998                                  * also create a new temporary address.
1999                                  */
2000                                 read_unlock_bh(&in6_dev->lock);
2001                                 ipv6_create_tempaddr(ifp, NULL);
2002                         } else {
2003                                 read_unlock_bh(&in6_dev->lock);
2004                         }
2005 #endif
2006                         in6_ifa_put(ifp);
2007                         addrconf_verify(0);
2008                 }
2009         }
2010         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2011         in6_dev_put(in6_dev);
2012 }
2013
2014 /*
2015  *      Set destination address.
2016  *      Special case for SIT interfaces where we create a new "virtual"
2017  *      device.
2018  */
2019 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2020 {
2021         struct in6_ifreq ireq;
2022         struct net_device *dev;
2023         int err = -EINVAL;
2024
2025         rtnl_lock();
2026
2027         err = -EFAULT;
2028         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2029                 goto err_exit;
2030
2031         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2032
2033         err = -ENODEV;
2034         if (dev == NULL)
2035                 goto err_exit;
2036
2037 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2038         if (dev->type == ARPHRD_SIT) {
2039                 const struct net_device_ops *ops = dev->netdev_ops;
2040                 struct ifreq ifr;
2041                 struct ip_tunnel_parm p;
2042
2043                 err = -EADDRNOTAVAIL;
2044                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2045                         goto err_exit;
2046
2047                 memset(&p, 0, sizeof(p));
2048                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2049                 p.iph.saddr = 0;
2050                 p.iph.version = 4;
2051                 p.iph.ihl = 5;
2052                 p.iph.protocol = IPPROTO_IPV6;
2053                 p.iph.ttl = 64;
2054                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2055
2056                 if (ops->ndo_do_ioctl) {
2057                         mm_segment_t oldfs = get_fs();
2058
2059                         set_fs(KERNEL_DS);
2060                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2061                         set_fs(oldfs);
2062                 } else
2063                         err = -EOPNOTSUPP;
2064
2065                 if (err == 0) {
2066                         err = -ENOBUFS;
2067                         dev = __dev_get_by_name(net, p.name);
2068                         if (!dev)
2069                                 goto err_exit;
2070                         err = dev_open(dev);
2071                 }
2072         }
2073 #endif
2074
2075 err_exit:
2076         rtnl_unlock();
2077         return err;
2078 }
2079
2080 /*
2081  *      Manual configuration of address on an interface
2082  */
2083 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2084                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2085                           __u32 valid_lft)
2086 {
2087         struct inet6_ifaddr *ifp;
2088         struct inet6_dev *idev;
2089         struct net_device *dev;
2090         int scope;
2091         u32 flags;
2092         clock_t expires;
2093         unsigned long timeout;
2094
2095         ASSERT_RTNL();
2096
2097         if (plen > 128)
2098                 return -EINVAL;
2099
2100         /* check the lifetime */
2101         if (!valid_lft || prefered_lft > valid_lft)
2102                 return -EINVAL;
2103
2104         dev = __dev_get_by_index(net, ifindex);
2105         if (!dev)
2106                 return -ENODEV;
2107
2108         if ((idev = addrconf_add_dev(dev)) == NULL)
2109                 return -ENOBUFS;
2110
2111         scope = ipv6_addr_scope(pfx);
2112
2113         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2114         if (addrconf_finite_timeout(timeout)) {
2115                 expires = jiffies_to_clock_t(timeout * HZ);
2116                 valid_lft = timeout;
2117                 flags = RTF_EXPIRES;
2118         } else {
2119                 expires = 0;
2120                 flags = 0;
2121                 ifa_flags |= IFA_F_PERMANENT;
2122         }
2123
2124         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2125         if (addrconf_finite_timeout(timeout)) {
2126                 if (timeout == 0)
2127                         ifa_flags |= IFA_F_DEPRECATED;
2128                 prefered_lft = timeout;
2129         }
2130
2131         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2132
2133         if (!IS_ERR(ifp)) {
2134                 spin_lock_bh(&ifp->lock);
2135                 ifp->valid_lft = valid_lft;
2136                 ifp->prefered_lft = prefered_lft;
2137                 ifp->tstamp = jiffies;
2138                 spin_unlock_bh(&ifp->lock);
2139
2140                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2141                                       expires, flags);
2142                 /*
2143                  * Note that section 3.1 of RFC 4429 indicates
2144                  * that the Optimistic flag should not be set for
2145                  * manually configured addresses
2146                  */
2147                 addrconf_dad_start(ifp, 0);
2148                 in6_ifa_put(ifp);
2149                 addrconf_verify(0);
2150                 return 0;
2151         }
2152
2153         return PTR_ERR(ifp);
2154 }
2155
2156 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2157                           unsigned int plen)
2158 {
2159         struct inet6_ifaddr *ifp;
2160         struct inet6_dev *idev;
2161         struct net_device *dev;
2162
2163         if (plen > 128)
2164                 return -EINVAL;
2165
2166         dev = __dev_get_by_index(net, ifindex);
2167         if (!dev)
2168                 return -ENODEV;
2169
2170         if ((idev = __in6_dev_get(dev)) == NULL)
2171                 return -ENXIO;
2172
2173         read_lock_bh(&idev->lock);
2174         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2175                 if (ifp->prefix_len == plen &&
2176                     ipv6_addr_equal(pfx, &ifp->addr)) {
2177                         in6_ifa_hold(ifp);
2178                         read_unlock_bh(&idev->lock);
2179
2180                         ipv6_del_addr(ifp);
2181
2182                         /* If the last address is deleted administratively,
2183                            disable IPv6 on this interface.
2184                          */
2185                         if (idev->addr_list == NULL)
2186                                 addrconf_ifdown(idev->dev, 1);
2187                         return 0;
2188                 }
2189         }
2190         read_unlock_bh(&idev->lock);
2191         return -EADDRNOTAVAIL;
2192 }
2193
2194
2195 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2196 {
2197         struct in6_ifreq ireq;
2198         int err;
2199
2200         if (!capable(CAP_NET_ADMIN))
2201                 return -EPERM;
2202
2203         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2204                 return -EFAULT;
2205
2206         rtnl_lock();
2207         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2208                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2209                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2210         rtnl_unlock();
2211         return err;
2212 }
2213
2214 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2215 {
2216         struct in6_ifreq ireq;
2217         int err;
2218
2219         if (!capable(CAP_NET_ADMIN))
2220                 return -EPERM;
2221
2222         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2223                 return -EFAULT;
2224
2225         rtnl_lock();
2226         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2227                              ireq.ifr6_prefixlen);
2228         rtnl_unlock();
2229         return err;
2230 }
2231
2232 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2233                      int plen, int scope)
2234 {
2235         struct inet6_ifaddr *ifp;
2236
2237         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2238         if (!IS_ERR(ifp)) {
2239                 spin_lock_bh(&ifp->lock);
2240                 ifp->flags &= ~IFA_F_TENTATIVE;
2241                 spin_unlock_bh(&ifp->lock);
2242                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2243                 in6_ifa_put(ifp);
2244         }
2245 }
2246
2247 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2248 static void sit_add_v4_addrs(struct inet6_dev *idev)
2249 {
2250         struct in6_addr addr;
2251         struct net_device *dev;
2252         struct net *net = dev_net(idev->dev);
2253         int scope;
2254
2255         ASSERT_RTNL();
2256
2257         memset(&addr, 0, sizeof(struct in6_addr));
2258         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2259
2260         if (idev->dev->flags&IFF_POINTOPOINT) {
2261                 addr.s6_addr32[0] = htonl(0xfe800000);
2262                 scope = IFA_LINK;
2263         } else {
2264                 scope = IPV6_ADDR_COMPATv4;
2265         }
2266
2267         if (addr.s6_addr32[3]) {
2268                 add_addr(idev, &addr, 128, scope);
2269                 return;
2270         }
2271
2272         for_each_netdev(net, dev) {
2273                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2274                 if (in_dev && (dev->flags & IFF_UP)) {
2275                         struct in_ifaddr * ifa;
2276
2277                         int flag = scope;
2278
2279                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2280                                 int plen;
2281
2282                                 addr.s6_addr32[3] = ifa->ifa_local;
2283
2284                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2285                                         continue;
2286                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2287                                         if (idev->dev->flags&IFF_POINTOPOINT)
2288                                                 continue;
2289                                         flag |= IFA_HOST;
2290                                 }
2291                                 if (idev->dev->flags&IFF_POINTOPOINT)
2292                                         plen = 64;
2293                                 else
2294                                         plen = 96;
2295
2296                                 add_addr(idev, &addr, plen, flag);
2297                         }
2298                 }
2299         }
2300 }
2301 #endif
2302
2303 static void init_loopback(struct net_device *dev)
2304 {
2305         struct inet6_dev  *idev;
2306
2307         /* ::1 */
2308
2309         ASSERT_RTNL();
2310
2311         if ((idev = ipv6_find_idev(dev)) == NULL) {
2312                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2313                 return;
2314         }
2315
2316         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2317 }
2318
2319 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2320 {
2321         struct inet6_ifaddr * ifp;
2322         u32 addr_flags = IFA_F_PERMANENT;
2323
2324 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2325         if (idev->cnf.optimistic_dad &&
2326             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2327                 addr_flags |= IFA_F_OPTIMISTIC;
2328 #endif
2329
2330
2331         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2332         if (!IS_ERR(ifp)) {
2333                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2334                 addrconf_dad_start(ifp, 0);
2335                 in6_ifa_put(ifp);
2336         }
2337 }
2338
2339 static void addrconf_dev_config(struct net_device *dev)
2340 {
2341         struct in6_addr addr;
2342         struct inet6_dev    * idev;
2343
2344         ASSERT_RTNL();
2345
2346         if ((dev->type != ARPHRD_ETHER) &&
2347             (dev->type != ARPHRD_FDDI) &&
2348             (dev->type != ARPHRD_IEEE802_TR) &&
2349             (dev->type != ARPHRD_ARCNET) &&
2350             (dev->type != ARPHRD_INFINIBAND)) {
2351                 /* Alas, we support only Ethernet autoconfiguration. */
2352                 return;
2353         }
2354
2355         idev = addrconf_add_dev(dev);
2356         if (idev == NULL)
2357                 return;
2358
2359         memset(&addr, 0, sizeof(struct in6_addr));
2360         addr.s6_addr32[0] = htonl(0xFE800000);
2361
2362         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2363                 addrconf_add_linklocal(idev, &addr);
2364 }
2365
2366 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2367 static void addrconf_sit_config(struct net_device *dev)
2368 {
2369         struct inet6_dev *idev;
2370
2371         ASSERT_RTNL();
2372
2373         /*
2374          * Configure the tunnel with one of our IPv4
2375          * addresses... we should configure all of
2376          * our v4 addrs in the tunnel
2377          */
2378
2379         if ((idev = ipv6_find_idev(dev)) == NULL) {
2380                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2381                 return;
2382         }
2383
2384         if (dev->priv_flags & IFF_ISATAP) {
2385                 struct in6_addr addr;
2386
2387                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2388                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2389                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2390                         addrconf_add_linklocal(idev, &addr);
2391                 return;
2392         }
2393
2394         sit_add_v4_addrs(idev);
2395
2396         if (dev->flags&IFF_POINTOPOINT) {
2397                 addrconf_add_mroute(dev);
2398                 addrconf_add_lroute(dev);
2399         } else
2400                 sit_route_add(dev);
2401 }
2402 #endif
2403
2404 static inline int
2405 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2406 {
2407         struct in6_addr lladdr;
2408
2409         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2410                 addrconf_add_linklocal(idev, &lladdr);
2411                 return 0;
2412         }
2413         return -1;
2414 }
2415
2416 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2417 {
2418         struct net_device *link_dev;
2419         struct net *net = dev_net(idev->dev);
2420
2421         /* first try to inherit the link-local address from the link device */
2422         if (idev->dev->iflink &&
2423             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2424                 if (!ipv6_inherit_linklocal(idev, link_dev))
2425                         return;
2426         }
2427         /* then try to inherit it from any device */
2428         for_each_netdev(net, link_dev) {
2429                 if (!ipv6_inherit_linklocal(idev, link_dev))
2430                         return;
2431         }
2432         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2433 }
2434
2435 /*
2436  * Autoconfigure tunnel with a link-local address so routing protocols,
2437  * DHCPv6, MLD etc. can be run over the virtual link
2438  */
2439
2440 static void addrconf_ip6_tnl_config(struct net_device *dev)
2441 {
2442         struct inet6_dev *idev;
2443
2444         ASSERT_RTNL();
2445
2446         if ((idev = addrconf_add_dev(dev)) == NULL) {
2447                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2448                 return;
2449         }
2450         ip6_tnl_add_linklocal(idev);
2451 }
2452
2453 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2454                            void * data)
2455 {
2456         struct net_device *dev = (struct net_device *) data;
2457         struct inet6_dev *idev = __in6_dev_get(dev);
2458         int run_pending = 0;
2459         int err;
2460
2461         switch(event) {
2462         case NETDEV_REGISTER:
2463                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2464                         idev = ipv6_add_dev(dev);
2465                         if (!idev)
2466                                 return notifier_from_errno(-ENOMEM);
2467                 }
2468                 break;
2469         case NETDEV_UP:
2470         case NETDEV_CHANGE:
2471                 if (dev->flags & IFF_SLAVE)
2472                         break;
2473
2474                 if (event == NETDEV_UP) {
2475                         if (!addrconf_qdisc_ok(dev)) {
2476                                 /* device is not ready yet. */
2477                                 printk(KERN_INFO
2478                                         "ADDRCONF(NETDEV_UP): %s: "
2479                                         "link is not ready\n",
2480                                         dev->name);
2481                                 break;
2482                         }
2483
2484                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2485                                 idev = ipv6_add_dev(dev);
2486
2487                         if (idev) {
2488                                 idev->if_flags |= IF_READY;
2489                                 run_pending = 1;
2490                         }
2491                 } else {
2492                         if (!addrconf_qdisc_ok(dev)) {
2493                                 /* device is still not ready. */
2494                                 break;
2495                         }
2496
2497                         if (idev) {
2498                                 if (idev->if_flags & IF_READY) {
2499                                         /* device is already configured. */
2500                                         break;
2501                                 }
2502                                 idev->if_flags |= IF_READY;
2503                         }
2504
2505                         printk(KERN_INFO
2506                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2507                                         "link becomes ready\n",
2508                                         dev->name);
2509
2510                         run_pending = 1;
2511                 }
2512
2513                 switch(dev->type) {
2514 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2515                 case ARPHRD_SIT:
2516                         addrconf_sit_config(dev);
2517                         break;
2518 #endif
2519                 case ARPHRD_TUNNEL6:
2520                         addrconf_ip6_tnl_config(dev);
2521                         break;
2522                 case ARPHRD_LOOPBACK:
2523                         init_loopback(dev);
2524                         break;
2525
2526                 default:
2527                         addrconf_dev_config(dev);
2528                         break;
2529                 }
2530                 if (idev) {
2531                         if (run_pending)
2532                                 addrconf_dad_run(idev);
2533
2534                         /* If the MTU changed during the interface down, when the
2535                            interface up, the changed MTU must be reflected in the
2536                            idev as well as routers.
2537                          */
2538                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2539                                 rt6_mtu_change(dev, dev->mtu);
2540                                 idev->cnf.mtu6 = dev->mtu;
2541                         }
2542                         idev->tstamp = jiffies;
2543                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2544                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2545                            stop IPv6 on this interface.
2546                          */
2547                         if (dev->mtu < IPV6_MIN_MTU)
2548                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2549                 }
2550                 break;
2551
2552         case NETDEV_CHANGEMTU:
2553                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2554                         rt6_mtu_change(dev, dev->mtu);
2555                         idev->cnf.mtu6 = dev->mtu;
2556                         break;
2557                 }
2558
2559                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2560                         idev = ipv6_add_dev(dev);
2561                         if (idev)
2562                                 break;
2563                 }
2564
2565                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2566
2567         case NETDEV_DOWN:
2568         case NETDEV_UNREGISTER:
2569                 /*
2570                  *      Remove all addresses from this interface.
2571                  */
2572                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2573                 break;
2574
2575         case NETDEV_CHANGENAME:
2576                 if (idev) {
2577                         snmp6_unregister_dev(idev);
2578                         addrconf_sysctl_unregister(idev);
2579                         addrconf_sysctl_register(idev);
2580                         err = snmp6_register_dev(idev);
2581                         if (err)
2582                                 return notifier_from_errno(err);
2583                 }
2584                 break;
2585         case NETDEV_BONDING_OLDTYPE:
2586         case NETDEV_BONDING_NEWTYPE:
2587                 addrconf_bonding_change(dev, event);
2588                 break;
2589         }
2590
2591         return NOTIFY_OK;
2592 }
2593
2594 /*
2595  *      addrconf module should be notified of a device going up
2596  */
2597 static struct notifier_block ipv6_dev_notf = {
2598         .notifier_call = addrconf_notify,
2599         .priority = 0
2600 };
2601
2602 static void addrconf_bonding_change(struct net_device *dev, unsigned long event)
2603 {
2604         struct inet6_dev *idev;
2605         ASSERT_RTNL();
2606
2607         idev = __in6_dev_get(dev);
2608
2609         if (event == NETDEV_BONDING_NEWTYPE)
2610                 ipv6_mc_remap(idev);
2611         else if (event == NETDEV_BONDING_OLDTYPE)
2612                 ipv6_mc_unmap(idev);
2613 }
2614
2615 static int addrconf_ifdown(struct net_device *dev, int how)
2616 {
2617         struct inet6_dev *idev;
2618         struct inet6_ifaddr *ifa, **bifa;
2619         struct net *net = dev_net(dev);
2620         int i;
2621
2622         ASSERT_RTNL();
2623
2624         rt6_ifdown(net, dev);
2625         neigh_ifdown(&nd_tbl, dev);
2626
2627         idev = __in6_dev_get(dev);
2628         if (idev == NULL)
2629                 return -ENODEV;
2630
2631         /* Step 1: remove reference to ipv6 device from parent device.
2632                    Do not dev_put!
2633          */
2634         if (how) {
2635                 idev->dead = 1;
2636
2637                 /* protected by rtnl_lock */
2638                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2639
2640                 /* Step 1.5: remove snmp6 entry */
2641                 snmp6_unregister_dev(idev);
2642
2643         }
2644
2645         /* Step 2: clear hash table */
2646         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2647                 bifa = &inet6_addr_lst[i];
2648
2649                 write_lock_bh(&addrconf_hash_lock);
2650                 while ((ifa = *bifa) != NULL) {
2651                         if (ifa->idev == idev &&
2652                             (how || !(ifa->flags&IFA_F_PERMANENT))) {
2653                                 *bifa = ifa->lst_next;
2654                                 ifa->lst_next = NULL;
2655                                 addrconf_del_timer(ifa);
2656                                 in6_ifa_put(ifa);
2657                                 continue;
2658                         }
2659                         bifa = &ifa->lst_next;
2660                 }
2661                 write_unlock_bh(&addrconf_hash_lock);
2662         }
2663
2664         write_lock_bh(&idev->lock);
2665
2666         /* Step 3: clear flags for stateless addrconf */
2667         if (!how)
2668                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2669
2670         /* Step 4: clear address list */
2671 #ifdef CONFIG_IPV6_PRIVACY
2672         if (how && del_timer(&idev->regen_timer))
2673                 in6_dev_put(idev);
2674
2675         /* clear tempaddr list */
2676         while ((ifa = idev->tempaddr_list) != NULL) {
2677                 idev->tempaddr_list = ifa->tmp_next;
2678                 ifa->tmp_next = NULL;
2679                 ifa->dead = 1;
2680                 write_unlock_bh(&idev->lock);
2681                 spin_lock_bh(&ifa->lock);
2682
2683                 if (ifa->ifpub) {
2684                         in6_ifa_put(ifa->ifpub);
2685                         ifa->ifpub = NULL;
2686                 }
2687                 spin_unlock_bh(&ifa->lock);
2688                 in6_ifa_put(ifa);
2689                 write_lock_bh(&idev->lock);
2690         }
2691 #endif
2692         bifa = &idev->addr_list;
2693         while ((ifa = *bifa) != NULL) {
2694                 if (how == 0 && (ifa->flags&IFA_F_PERMANENT)) {
2695                         /* Retain permanent address on admin down */
2696                         bifa = &ifa->if_next;
2697
2698                         /* Restart DAD if needed when link comes back up */
2699                         if ( !((dev->flags&(IFF_NOARP|IFF_LOOPBACK)) ||
2700                                idev->cnf.accept_dad <= 0 ||
2701                                (ifa->flags & IFA_F_NODAD)))
2702                                 ifa->flags |= IFA_F_TENTATIVE;
2703                 } else {
2704                         *bifa = ifa->if_next;
2705                         ifa->if_next = NULL;
2706
2707                         ifa->dead = 1;
2708                         write_unlock_bh(&idev->lock);
2709
2710                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2711                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2712                         in6_ifa_put(ifa);
2713
2714                         write_lock_bh(&idev->lock);
2715                 }
2716         }
2717         write_unlock_bh(&idev->lock);
2718
2719         /* Step 5: Discard multicast list */
2720
2721         if (how)
2722                 ipv6_mc_destroy_dev(idev);
2723         else
2724                 ipv6_mc_down(idev);
2725
2726         idev->tstamp = jiffies;
2727
2728         /* Shot the device (if unregistered) */
2729
2730         if (how) {
2731                 addrconf_sysctl_unregister(idev);
2732                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2733                 neigh_ifdown(&nd_tbl, dev);
2734                 in6_dev_put(idev);
2735         }
2736         return 0;
2737 }
2738
2739 static void addrconf_rs_timer(unsigned long data)
2740 {
2741         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2742
2743         if (ifp->idev->cnf.forwarding)
2744                 goto out;
2745
2746         if (ifp->idev->if_flags & IF_RA_RCVD) {
2747                 /*
2748                  *      Announcement received after solicitation
2749                  *      was sent
2750                  */
2751                 goto out;
2752         }
2753
2754         spin_lock(&ifp->lock);
2755         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2756                 /* The wait after the last probe can be shorter */
2757                 addrconf_mod_timer(ifp, AC_RS,
2758                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2759                                    ifp->idev->cnf.rtr_solicit_delay :
2760                                    ifp->idev->cnf.rtr_solicit_interval);
2761                 spin_unlock(&ifp->lock);
2762
2763                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2764         } else {
2765                 spin_unlock(&ifp->lock);
2766                 /*
2767                  * Note: we do not support deprecated "all on-link"
2768                  * assumption any longer.
2769                  */
2770                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2771                        ifp->idev->dev->name);
2772         }
2773
2774 out:
2775         in6_ifa_put(ifp);
2776 }
2777
2778 /*
2779  *      Duplicate Address Detection
2780  */
2781 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2782 {
2783         unsigned long rand_num;
2784         struct inet6_dev *idev = ifp->idev;
2785
2786         if (ifp->flags & IFA_F_OPTIMISTIC)
2787                 rand_num = 0;
2788         else
2789                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2790
2791         ifp->probes = idev->cnf.dad_transmits;
2792         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2793 }
2794
2795 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2796 {
2797         struct inet6_dev *idev = ifp->idev;
2798         struct net_device *dev = idev->dev;
2799
2800         addrconf_join_solict(dev, &ifp->addr);
2801
2802         net_srandom(ifp->addr.s6_addr32[3]);
2803
2804         read_lock_bh(&idev->lock);
2805         if (ifp->dead)
2806                 goto out;
2807
2808         spin_lock(&ifp->lock);
2809         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2810             idev->cnf.accept_dad < 1 ||
2811             !(ifp->flags&IFA_F_TENTATIVE) ||
2812             ifp->flags & IFA_F_NODAD) {
2813                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2814                 spin_unlock(&ifp->lock);
2815                 read_unlock_bh(&idev->lock);
2816
2817                 addrconf_dad_completed(ifp);
2818                 return;
2819         }
2820
2821         if (!(idev->if_flags & IF_READY)) {
2822                 spin_unlock(&ifp->lock);
2823                 read_unlock_bh(&idev->lock);
2824                 /*
2825                  * If the device is not ready:
2826                  * - keep it tentative if it is a permanent address.
2827                  * - otherwise, kill it.
2828                  */
2829                 in6_ifa_hold(ifp);
2830                 addrconf_dad_stop(ifp, 0);
2831                 return;
2832         }
2833
2834         /*
2835          * Optimistic nodes can start receiving
2836          * Frames right away
2837          */
2838         if(ifp->flags & IFA_F_OPTIMISTIC)
2839                 ip6_ins_rt(ifp->rt);
2840
2841         addrconf_dad_kick(ifp);
2842         spin_unlock(&ifp->lock);
2843 out:
2844         read_unlock_bh(&idev->lock);
2845 }
2846
2847 static void addrconf_dad_timer(unsigned long data)
2848 {
2849         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2850         struct inet6_dev *idev = ifp->idev;
2851         struct in6_addr mcaddr;
2852
2853         read_lock_bh(&idev->lock);
2854         if (idev->dead) {
2855                 read_unlock_bh(&idev->lock);
2856                 goto out;
2857         }
2858
2859         spin_lock(&ifp->lock);
2860         if (ifp->probes == 0) {
2861                 /*
2862                  * DAD was successful
2863                  */
2864
2865                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2866                 spin_unlock(&ifp->lock);
2867                 read_unlock_bh(&idev->lock);
2868
2869                 addrconf_dad_completed(ifp);
2870
2871                 goto out;
2872         }
2873
2874         ifp->probes--;
2875         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2876         spin_unlock(&ifp->lock);
2877         read_unlock_bh(&idev->lock);
2878
2879         /* send a neighbour solicitation for our addr */
2880         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2881         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2882 out:
2883         in6_ifa_put(ifp);
2884 }
2885
2886 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2887 {
2888         struct net_device *     dev = ifp->idev->dev;
2889
2890         /*
2891          *      Configure the address for reception. Now it is valid.
2892          */
2893
2894         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2895
2896         /* If added prefix is link local and forwarding is off,
2897            start sending router solicitations.
2898          */
2899
2900         if (ifp->idev->cnf.forwarding == 0 &&
2901             ifp->idev->cnf.rtr_solicits > 0 &&
2902             (dev->flags&IFF_LOOPBACK) == 0 &&
2903             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2904                 /*
2905                  *      If a host as already performed a random delay
2906                  *      [...] as part of DAD [...] there is no need
2907                  *      to delay again before sending the first RS
2908                  */
2909                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2910
2911                 spin_lock_bh(&ifp->lock);
2912                 ifp->probes = 1;
2913                 ifp->idev->if_flags |= IF_RS_SENT;
2914                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2915                 spin_unlock_bh(&ifp->lock);
2916         }
2917 }
2918
2919 static void addrconf_dad_run(struct inet6_dev *idev) {
2920         struct inet6_ifaddr *ifp;
2921
2922         read_lock_bh(&idev->lock);
2923         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2924                 spin_lock(&ifp->lock);
2925                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2926                         spin_unlock(&ifp->lock);
2927                         continue;
2928                 }
2929                 spin_unlock(&ifp->lock);
2930                 addrconf_dad_kick(ifp);
2931         }
2932         read_unlock_bh(&idev->lock);
2933 }
2934
2935 #ifdef CONFIG_PROC_FS
2936 struct if6_iter_state {
2937         struct seq_net_private p;
2938         int bucket;
2939 };
2940
2941 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2942 {
2943         struct inet6_ifaddr *ifa = NULL;
2944         struct if6_iter_state *state = seq->private;
2945         struct net *net = seq_file_net(seq);
2946
2947         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2948                 ifa = inet6_addr_lst[state->bucket];
2949
2950                 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2951                         ifa = ifa->lst_next;
2952                 if (ifa)
2953                         break;
2954         }
2955         return ifa;
2956 }
2957
2958 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2959 {
2960         struct if6_iter_state *state = seq->private;
2961         struct net *net = seq_file_net(seq);
2962
2963         ifa = ifa->lst_next;
2964 try_again:
2965         if (ifa) {
2966                 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2967                         ifa = ifa->lst_next;
2968                         goto try_again;
2969                 }
2970         }
2971
2972         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2973                 ifa = inet6_addr_lst[state->bucket];
2974                 goto try_again;
2975         }
2976
2977         return ifa;
2978 }
2979
2980 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2981 {
2982         struct inet6_ifaddr *ifa = if6_get_first(seq);
2983
2984         if (ifa)
2985                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2986                         --pos;
2987         return pos ? NULL : ifa;
2988 }
2989
2990 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2991         __acquires(addrconf_hash_lock)
2992 {
2993         read_lock_bh(&addrconf_hash_lock);
2994         return if6_get_idx(seq, *pos);
2995 }
2996
2997 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2998 {
2999         struct inet6_ifaddr *ifa;
3000
3001         ifa = if6_get_next(seq, v);
3002         ++*pos;
3003         return ifa;
3004 }
3005
3006 static void if6_seq_stop(struct seq_file *seq, void *v)
3007         __releases(addrconf_hash_lock)
3008 {
3009         read_unlock_bh(&addrconf_hash_lock);
3010 }
3011
3012 static int if6_seq_show(struct seq_file *seq, void *v)
3013 {
3014         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3015         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3016                    &ifp->addr,
3017                    ifp->idev->dev->ifindex,
3018                    ifp->prefix_len,
3019                    ifp->scope,
3020                    ifp->flags,
3021                    ifp->idev->dev->name);
3022         return 0;
3023 }
3024
3025 static const struct seq_operations if6_seq_ops = {
3026         .start  = if6_seq_start,
3027         .next   = if6_seq_next,
3028         .show   = if6_seq_show,
3029         .stop   = if6_seq_stop,
3030 };
3031
3032 static int if6_seq_open(struct inode *inode, struct file *file)
3033 {
3034         return seq_open_net(inode, file, &if6_seq_ops,
3035                             sizeof(struct if6_iter_state));
3036 }
3037
3038 static const struct file_operations if6_fops = {
3039         .owner          = THIS_MODULE,
3040         .open           = if6_seq_open,
3041         .read           = seq_read,
3042         .llseek         = seq_lseek,
3043         .release        = seq_release_net,
3044 };
3045
3046 static int __net_init if6_proc_net_init(struct net *net)
3047 {
3048         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3049                 return -ENOMEM;
3050         return 0;
3051 }
3052
3053 static void __net_exit if6_proc_net_exit(struct net *net)
3054 {
3055        proc_net_remove(net, "if_inet6");
3056 }
3057
3058 static struct pernet_operations if6_proc_net_ops = {
3059        .init = if6_proc_net_init,
3060        .exit = if6_proc_net_exit,
3061 };
3062
3063 int __init if6_proc_init(void)
3064 {
3065         return register_pernet_subsys(&if6_proc_net_ops);
3066 }
3067
3068 void if6_proc_exit(void)
3069 {
3070         unregister_pernet_subsys(&if6_proc_net_ops);
3071 }
3072 #endif  /* CONFIG_PROC_FS */
3073
3074 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3075 /* Check if address is a home address configured on any interface. */
3076 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3077 {
3078         int ret = 0;
3079         struct inet6_ifaddr * ifp;
3080         u8 hash = ipv6_addr_hash(addr);
3081         read_lock_bh(&addrconf_hash_lock);
3082         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
3083                 if (!net_eq(dev_net(ifp->idev->dev), net))
3084                         continue;
3085                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3086                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3087                         ret = 1;
3088                         break;
3089                 }
3090         }
3091         read_unlock_bh(&addrconf_hash_lock);
3092         return ret;
3093 }
3094 #endif
3095
3096 /*
3097  *      Periodic address status verification
3098  */
3099
3100 static void addrconf_verify(unsigned long foo)
3101 {
3102         struct inet6_ifaddr *ifp;
3103         unsigned long now, next;
3104         int i;
3105
3106         spin_lock_bh(&addrconf_verify_lock);
3107         now = jiffies;
3108         next = now + ADDR_CHECK_FREQUENCY;
3109
3110         del_timer(&addr_chk_timer);
3111
3112         for (i=0; i < IN6_ADDR_HSIZE; i++) {
3113
3114 restart:
3115                 read_lock(&addrconf_hash_lock);
3116                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3117                         unsigned long age;
3118 #ifdef CONFIG_IPV6_PRIVACY
3119                         unsigned long regen_advance;
3120 #endif
3121
3122                         if (ifp->flags & IFA_F_PERMANENT)
3123                                 continue;
3124
3125                         spin_lock(&ifp->lock);
3126                         age = (now - ifp->tstamp) / HZ;
3127
3128 #ifdef CONFIG_IPV6_PRIVACY
3129                         regen_advance = ifp->idev->cnf.regen_max_retry *
3130                                         ifp->idev->cnf.dad_transmits *
3131                                         ifp->idev->nd_parms->retrans_time / HZ;
3132 #endif
3133
3134                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3135                             age >= ifp->valid_lft) {
3136                                 spin_unlock(&ifp->lock);
3137                                 in6_ifa_hold(ifp);
3138                                 read_unlock(&addrconf_hash_lock);
3139                                 ipv6_del_addr(ifp);
3140                                 goto restart;
3141                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3142                                 spin_unlock(&ifp->lock);
3143                                 continue;
3144                         } else if (age >= ifp->prefered_lft) {
3145                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3146                                 int deprecate = 0;
3147
3148                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3149                                         deprecate = 1;
3150                                         ifp->flags |= IFA_F_DEPRECATED;
3151                                 }
3152
3153                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3154                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3155
3156                                 spin_unlock(&ifp->lock);
3157
3158                                 if (deprecate) {
3159                                         in6_ifa_hold(ifp);
3160                                         read_unlock(&addrconf_hash_lock);
3161
3162                                         ipv6_ifa_notify(0, ifp);
3163                                         in6_ifa_put(ifp);
3164                                         goto restart;
3165                                 }
3166 #ifdef CONFIG_IPV6_PRIVACY
3167                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3168                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3169                                 if (age >= ifp->prefered_lft - regen_advance) {
3170                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3171                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3172                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3173                                         if (!ifp->regen_count && ifpub) {
3174                                                 ifp->regen_count++;
3175                                                 in6_ifa_hold(ifp);
3176                                                 in6_ifa_hold(ifpub);
3177                                                 spin_unlock(&ifp->lock);
3178                                                 read_unlock(&addrconf_hash_lock);
3179                                                 spin_lock(&ifpub->lock);
3180                                                 ifpub->regen_count = 0;
3181                                                 spin_unlock(&ifpub->lock);
3182                                                 ipv6_create_tempaddr(ifpub, ifp);
3183                                                 in6_ifa_put(ifpub);
3184                                                 in6_ifa_put(ifp);
3185                                                 goto restart;
3186                                         }
3187                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3188                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3189                                 spin_unlock(&ifp->lock);
3190 #endif
3191                         } else {
3192                                 /* ifp->prefered_lft <= ifp->valid_lft */
3193                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3194                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3195                                 spin_unlock(&ifp->lock);
3196                         }
3197                 }
3198                 read_unlock(&addrconf_hash_lock);
3199         }
3200
3201         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3202         add_timer(&addr_chk_timer);
3203         spin_unlock_bh(&addrconf_verify_lock);
3204 }
3205
3206 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3207 {
3208         struct in6_addr *pfx = NULL;
3209
3210         if (addr)
3211                 pfx = nla_data(addr);
3212
3213         if (local) {
3214                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3215                         pfx = NULL;
3216                 else
3217                         pfx = nla_data(local);
3218         }
3219
3220         return pfx;
3221 }
3222
3223 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3224         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3225         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3226         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3227 };
3228
3229 static int
3230 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3231 {
3232         struct net *net = sock_net(skb->sk);
3233         struct ifaddrmsg *ifm;
3234         struct nlattr *tb[IFA_MAX+1];
3235         struct in6_addr *pfx;
3236         int err;
3237
3238         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3239         if (err < 0)
3240                 return err;
3241
3242         ifm = nlmsg_data(nlh);
3243         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3244         if (pfx == NULL)
3245                 return -EINVAL;
3246
3247         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3248 }
3249
3250 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3251                              u32 prefered_lft, u32 valid_lft)
3252 {
3253         u32 flags;
3254         clock_t expires;
3255         unsigned long timeout;
3256
3257         if (!valid_lft || (prefered_lft > valid_lft))
3258                 return -EINVAL;
3259
3260         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3261         if (addrconf_finite_timeout(timeout)) {
3262                 expires = jiffies_to_clock_t(timeout * HZ);
3263                 valid_lft = timeout;
3264                 flags = RTF_EXPIRES;
3265         } else {
3266                 expires = 0;
3267                 flags = 0;
3268                 ifa_flags |= IFA_F_PERMANENT;
3269         }
3270
3271         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3272         if (addrconf_finite_timeout(timeout)) {
3273                 if (timeout == 0)
3274                         ifa_flags |= IFA_F_DEPRECATED;
3275                 prefered_lft = timeout;
3276         }
3277
3278         spin_lock_bh(&ifp->lock);
3279         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3280         ifp->tstamp = jiffies;
3281         ifp->valid_lft = valid_lft;
3282         ifp->prefered_lft = prefered_lft;
3283
3284         spin_unlock_bh(&ifp->lock);
3285         if (!(ifp->flags&IFA_F_TENTATIVE))
3286                 ipv6_ifa_notify(0, ifp);
3287
3288         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3289                               expires, flags);
3290         addrconf_verify(0);
3291
3292         return 0;
3293 }
3294
3295 static int
3296 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3297 {
3298         struct net *net = sock_net(skb->sk);
3299         struct ifaddrmsg *ifm;
3300         struct nlattr *tb[IFA_MAX+1];
3301         struct in6_addr *pfx;
3302         struct inet6_ifaddr *ifa;
3303         struct net_device *dev;
3304         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3305         u8 ifa_flags;
3306         int err;
3307
3308         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3309         if (err < 0)
3310                 return err;
3311
3312         ifm = nlmsg_data(nlh);
3313         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3314         if (pfx == NULL)
3315                 return -EINVAL;
3316
3317         if (tb[IFA_CACHEINFO]) {
3318                 struct ifa_cacheinfo *ci;
3319
3320                 ci = nla_data(tb[IFA_CACHEINFO]);
3321                 valid_lft = ci->ifa_valid;
3322                 preferred_lft = ci->ifa_prefered;
3323         } else {
3324                 preferred_lft = INFINITY_LIFE_TIME;
3325                 valid_lft = INFINITY_LIFE_TIME;
3326         }
3327
3328         dev =  __dev_get_by_index(net, ifm->ifa_index);
3329         if (dev == NULL)
3330                 return -ENODEV;
3331
3332         /* We ignore other flags so far. */
3333         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3334
3335         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3336         if (ifa == NULL) {
3337                 /*
3338                  * It would be best to check for !NLM_F_CREATE here but
3339                  * userspace alreay relies on not having to provide this.
3340                  */
3341                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3342                                       ifm->ifa_prefixlen, ifa_flags,
3343                                       preferred_lft, valid_lft);
3344         }
3345
3346         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3347             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3348                 err = -EEXIST;
3349         else
3350                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3351
3352         in6_ifa_put(ifa);
3353
3354         return err;
3355 }
3356
3357 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3358                           u8 scope, int ifindex)
3359 {
3360         struct ifaddrmsg *ifm;
3361
3362         ifm = nlmsg_data(nlh);
3363         ifm->ifa_family = AF_INET6;
3364         ifm->ifa_prefixlen = prefixlen;
3365         ifm->ifa_flags = flags;
3366         ifm->ifa_scope = scope;
3367         ifm->ifa_index = ifindex;
3368 }
3369
3370 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3371                          unsigned long tstamp, u32 preferred, u32 valid)
3372 {
3373         struct ifa_cacheinfo ci;
3374
3375         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3376                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3377         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3378                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3379         ci.ifa_prefered = preferred;
3380         ci.ifa_valid = valid;
3381
3382         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3383 }
3384
3385 static inline int rt_scope(int ifa_scope)
3386 {
3387         if (ifa_scope & IFA_HOST)
3388                 return RT_SCOPE_HOST;
3389         else if (ifa_scope & IFA_LINK)
3390                 return RT_SCOPE_LINK;
3391         else if (ifa_scope & IFA_SITE)
3392                 return RT_SCOPE_SITE;
3393         else
3394                 return RT_SCOPE_UNIVERSE;
3395 }
3396
3397 static inline int inet6_ifaddr_msgsize(void)
3398 {
3399         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3400                + nla_total_size(16) /* IFA_ADDRESS */
3401                + nla_total_size(sizeof(struct ifa_cacheinfo));
3402 }
3403
3404 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3405                              u32 pid, u32 seq, int event, unsigned int flags)
3406 {
3407         struct nlmsghdr  *nlh;
3408         u32 preferred, valid;
3409
3410         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3411         if (nlh == NULL)
3412                 return -EMSGSIZE;
3413
3414         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3415                       ifa->idev->dev->ifindex);
3416
3417         if (!(ifa->flags&IFA_F_PERMANENT)) {
3418                 preferred = ifa->prefered_lft;
3419                 valid = ifa->valid_lft;
3420                 if (preferred != INFINITY_LIFE_TIME) {
3421                         long tval = (jiffies - ifa->tstamp)/HZ;
3422                         if (preferred > tval)
3423                                 preferred -= tval;
3424                         else
3425                                 preferred = 0;
3426                         if (valid != INFINITY_LIFE_TIME)
3427                                 valid -= tval;
3428                 }
3429         } else {
3430                 preferred = INFINITY_LIFE_TIME;
3431                 valid = INFINITY_LIFE_TIME;
3432         }
3433
3434         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3435             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3436                 nlmsg_cancel(skb, nlh);
3437                 return -EMSGSIZE;
3438         }
3439
3440         return nlmsg_end(skb, nlh);
3441 }
3442
3443 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3444                                 u32 pid, u32 seq, int event, u16 flags)
3445 {
3446         struct nlmsghdr  *nlh;
3447         u8 scope = RT_SCOPE_UNIVERSE;
3448         int ifindex = ifmca->idev->dev->ifindex;
3449
3450         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3451                 scope = RT_SCOPE_SITE;
3452
3453         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3454         if (nlh == NULL)
3455                 return -EMSGSIZE;
3456
3457         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3458         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3459             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3460                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3461                 nlmsg_cancel(skb, nlh);
3462                 return -EMSGSIZE;
3463         }
3464
3465         return nlmsg_end(skb, nlh);
3466 }
3467
3468 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3469                                 u32 pid, u32 seq, int event, unsigned int flags)
3470 {
3471         struct nlmsghdr  *nlh;
3472         u8 scope = RT_SCOPE_UNIVERSE;
3473         int ifindex = ifaca->aca_idev->dev->ifindex;
3474
3475         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3476                 scope = RT_SCOPE_SITE;
3477
3478         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3479         if (nlh == NULL)
3480                 return -EMSGSIZE;
3481
3482         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3483         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3484             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3485                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3486                 nlmsg_cancel(skb, nlh);
3487                 return -EMSGSIZE;
3488         }
3489
3490         return nlmsg_end(skb, nlh);
3491 }
3492
3493 enum addr_type_t
3494 {
3495         UNICAST_ADDR,
3496         MULTICAST_ADDR,
3497         ANYCAST_ADDR,
3498 };
3499
3500 /* called with rcu_read_lock() */
3501 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3502                           struct netlink_callback *cb, enum addr_type_t type,
3503                           int s_ip_idx, int *p_ip_idx)
3504 {
3505         struct inet6_ifaddr *ifa;
3506         struct ifmcaddr6 *ifmca;
3507         struct ifacaddr6 *ifaca;
3508         int err = 1;
3509         int ip_idx = *p_ip_idx;
3510
3511         read_lock_bh(&idev->lock);
3512         switch (type) {
3513         case UNICAST_ADDR:
3514                 /* unicast address incl. temp addr */
3515                 for (ifa = idev->addr_list; ifa;
3516                      ifa = ifa->if_next, ip_idx++) {
3517                         if (ip_idx < s_ip_idx)
3518                                 continue;
3519                         err = inet6_fill_ifaddr(skb, ifa,
3520                                                 NETLINK_CB(cb->skb).pid,
3521                                                 cb->nlh->nlmsg_seq,
3522                                                 RTM_NEWADDR,
3523                                                 NLM_F_MULTI);
3524                         if (err <= 0)
3525                                 break;
3526                 }
3527                 break;
3528         case MULTICAST_ADDR:
3529                 /* multicast address */
3530                 for (ifmca = idev->mc_list; ifmca;
3531                      ifmca = ifmca->next, ip_idx++) {
3532                         if (ip_idx < s_ip_idx)
3533                                 continue;
3534                         err = inet6_fill_ifmcaddr(skb, ifmca,
3535                                                   NETLINK_CB(cb->skb).pid,
3536                                                   cb->nlh->nlmsg_seq,
3537                                                   RTM_GETMULTICAST,
3538                                                   NLM_F_MULTI);
3539                         if (err <= 0)
3540                                 break;
3541                 }
3542                 break;
3543         case ANYCAST_ADDR:
3544                 /* anycast address */
3545                 for (ifaca = idev->ac_list; ifaca;
3546                      ifaca = ifaca->aca_next, ip_idx++) {
3547                         if (ip_idx < s_ip_idx)
3548                                 continue;
3549                         err = inet6_fill_ifacaddr(skb, ifaca,
3550                                                   NETLINK_CB(cb->skb).pid,
3551                                                   cb->nlh->nlmsg_seq,
3552                                                   RTM_GETANYCAST,
3553                                                   NLM_F_MULTI);
3554                         if (err <= 0)
3555                                 break;
3556                 }
3557                 break;
3558         default:
3559                 break;
3560         }
3561         read_unlock_bh(&idev->lock);
3562         *p_ip_idx = ip_idx;
3563         return err;
3564 }
3565
3566 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3567                            enum addr_type_t type)
3568 {
3569         struct net *net = sock_net(skb->sk);
3570         int h, s_h;
3571         int idx, ip_idx;
3572         int s_idx, s_ip_idx;
3573         struct net_device *dev;
3574         struct inet6_dev *idev;
3575         struct hlist_head *head;
3576         struct hlist_node *node;
3577
3578         s_h = cb->args[0];
3579         s_idx = idx = cb->args[1];
3580         s_ip_idx = ip_idx = cb->args[2];
3581
3582         rcu_read_lock();
3583         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3584                 idx = 0;
3585                 head = &net->dev_index_head[h];
3586                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3587                         if (idx < s_idx)
3588                                 goto cont;
3589                         if (idx > s_idx)
3590                                 s_ip_idx = 0;
3591                         ip_idx = 0;
3592                         if ((idev = __in6_dev_get(dev)) == NULL)
3593                                 goto cont;
3594
3595                         if (in6_dump_addrs(idev, skb, cb, type,
3596                                            s_ip_idx, &ip_idx) <= 0)
3597                                 goto done;
3598 cont:
3599                         idx++;
3600                 }
3601         }
3602 done:
3603         rcu_read_unlock();
3604         cb->args[0] = h;
3605         cb->args[1] = idx;
3606         cb->args[2] = ip_idx;
3607
3608         return skb->len;
3609 }
3610
3611 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3612 {
3613         enum addr_type_t type = UNICAST_ADDR;
3614
3615         return inet6_dump_addr(skb, cb, type);
3616 }
3617
3618 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3619 {
3620         enum addr_type_t type = MULTICAST_ADDR;
3621
3622         return inet6_dump_addr(skb, cb, type);
3623 }
3624
3625
3626 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3627 {
3628         enum addr_type_t type = ANYCAST_ADDR;
3629
3630         return inet6_dump_addr(skb, cb, type);
3631 }
3632
3633 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3634                              void *arg)
3635 {
3636         struct net *net = sock_net(in_skb->sk);
3637         struct ifaddrmsg *ifm;
3638         struct nlattr *tb[IFA_MAX+1];
3639         struct in6_addr *addr = NULL;
3640         struct net_device *dev = NULL;
3641         struct inet6_ifaddr *ifa;
3642         struct sk_buff *skb;
3643         int err;
3644
3645         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3646         if (err < 0)
3647                 goto errout;
3648
3649         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3650         if (addr == NULL) {
3651                 err = -EINVAL;
3652                 goto errout;
3653         }
3654
3655         ifm = nlmsg_data(nlh);
3656         if (ifm->ifa_index)
3657                 dev = __dev_get_by_index(net, ifm->ifa_index);
3658
3659         if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3660                 err = -EADDRNOTAVAIL;
3661                 goto errout;
3662         }
3663
3664         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3665                 err = -ENOBUFS;
3666                 goto errout_ifa;
3667         }
3668
3669         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3670                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3671         if (err < 0) {
3672                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3673                 WARN_ON(err == -EMSGSIZE);
3674                 kfree_skb(skb);
3675                 goto errout_ifa;
3676         }
3677         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3678 errout_ifa:
3679         in6_ifa_put(ifa);
3680 errout:
3681         return err;
3682 }
3683
3684 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3685 {
3686         struct sk_buff *skb;
3687         struct net *net = dev_net(ifa->idev->dev);
3688         int err = -ENOBUFS;
3689
3690         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3691         if (skb == NULL)
3692                 goto errout;
3693
3694         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3695         if (err < 0) {
3696                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3697                 WARN_ON(err == -EMSGSIZE);
3698                 kfree_skb(skb);
3699                 goto errout;
3700         }
3701         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3702         return;
3703 errout:
3704         if (err < 0)
3705                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3706 }
3707
3708 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3709                                 __s32 *array, int bytes)
3710 {
3711         BUG_ON(bytes < (DEVCONF_MAX * 4));
3712
3713         memset(array, 0, bytes);
3714         array[DEVCONF_FORWARDING] = cnf->forwarding;
3715         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3716         array[DEVCONF_MTU6] = cnf->mtu6;
3717         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3718         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3719         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3720         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3721         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3722         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3723         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3724         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3725 #ifdef CONFIG_IPV6_PRIVACY
3726         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3727         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3728         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3729         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3730         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3731 #endif
3732         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3733         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3734         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3735 #ifdef CONFIG_IPV6_ROUTER_PREF
3736         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3737         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3738 #ifdef CONFIG_IPV6_ROUTE_INFO
3739         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3740 #endif
3741 #endif
3742         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3743         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3744 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3745         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3746 #endif
3747 #ifdef CONFIG_IPV6_MROUTE
3748         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3749 #endif
3750         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3751         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3752         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3753 }
3754
3755 static inline size_t inet6_if_nlmsg_size(void)
3756 {
3757         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3758                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3759                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3760                + nla_total_size(4) /* IFLA_MTU */
3761                + nla_total_size(4) /* IFLA_LINK */
3762                + nla_total_size( /* IFLA_PROTINFO */
3763                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3764                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3765                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3766                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3767                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3768                  );
3769 }
3770
3771 static inline void __snmp6_fill_stats(u64 *stats, void __percpu **mib,
3772                                       int items, int bytes)
3773 {
3774         int i;
3775         int pad = bytes - sizeof(u64) * items;
3776         BUG_ON(pad < 0);
3777
3778         /* Use put_unaligned() because stats may not be aligned for u64. */
3779         put_unaligned(items, &stats[0]);
3780         for (i = 1; i < items; i++)
3781                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3782
3783         memset(&stats[items], 0, pad);
3784 }
3785
3786 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3787                              int bytes)
3788 {
3789         switch(attrtype) {
3790         case IFLA_INET6_STATS:
3791                 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3792                 break;
3793         case IFLA_INET6_ICMP6STATS:
3794                 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3795                 break;
3796         }
3797 }
3798
3799 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3800                              u32 pid, u32 seq, int event, unsigned int flags)
3801 {
3802         struct net_device *dev = idev->dev;
3803         struct nlattr *nla;
3804         struct ifinfomsg *hdr;
3805         struct nlmsghdr *nlh;
3806         void *protoinfo;
3807         struct ifla_cacheinfo ci;
3808
3809         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3810         if (nlh == NULL)
3811                 return -EMSGSIZE;
3812
3813         hdr = nlmsg_data(nlh);
3814         hdr->ifi_family = AF_INET6;
3815         hdr->__ifi_pad = 0;
3816         hdr->ifi_type = dev->type;
3817         hdr->ifi_index = dev->ifindex;
3818         hdr->ifi_flags = dev_get_flags(dev);
3819         hdr->ifi_change = 0;
3820
3821         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3822
3823         if (dev->addr_len)
3824                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3825
3826         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3827         if (dev->ifindex != dev->iflink)
3828                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3829
3830         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3831         if (protoinfo == NULL)
3832                 goto nla_put_failure;
3833
3834         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3835
3836         ci.max_reasm_len = IPV6_MAXPLEN;
3837         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3838                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3839         ci.reachable_time = idev->nd_parms->reachable_time;
3840         ci.retrans_time = idev->nd_parms->retrans_time;
3841         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3842
3843         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3844         if (nla == NULL)
3845                 goto nla_put_failure;
3846         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3847
3848         /* XXX - MC not implemented */
3849
3850         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3851         if (nla == NULL)
3852                 goto nla_put_failure;
3853         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3854
3855         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3856         if (nla == NULL)
3857                 goto nla_put_failure;
3858         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3859
3860         nla_nest_end(skb, protoinfo);
3861         return nlmsg_end(skb, nlh);
3862
3863 nla_put_failure:
3864         nlmsg_cancel(skb, nlh);
3865         return -EMSGSIZE;
3866 }
3867
3868 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3869 {
3870         struct net *net = sock_net(skb->sk);
3871         int h, s_h;
3872         int idx = 0, s_idx;
3873         struct net_device *dev;
3874         struct inet6_dev *idev;
3875         struct hlist_head *head;
3876         struct hlist_node *node;
3877
3878         s_h = cb->args[0];
3879         s_idx = cb->args[1];
3880
3881         rcu_read_lock();
3882         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3883                 idx = 0;
3884                 head = &net->dev_index_head[h];
3885                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3886                         if (idx < s_idx)
3887                                 goto cont;
3888                         idev = __in6_dev_get(dev);
3889                         if (!idev)
3890                                 goto cont;
3891                         if (inet6_fill_ifinfo(skb, idev,
3892                                               NETLINK_CB(cb->skb).pid,
3893                                               cb->nlh->nlmsg_seq,
3894                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
3895                                 goto out;
3896 cont:
3897                         idx++;
3898                 }
3899         }
3900 out:
3901         rcu_read_unlock();
3902         cb->args[1] = idx;
3903         cb->args[0] = h;
3904
3905         return skb->len;
3906 }
3907
3908 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3909 {
3910         struct sk_buff *skb;
3911         struct net *net = dev_net(idev->dev);
3912         int err = -ENOBUFS;
3913
3914         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3915         if (skb == NULL)
3916                 goto errout;
3917
3918         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3919         if (err < 0) {
3920                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3921                 WARN_ON(err == -EMSGSIZE);
3922                 kfree_skb(skb);
3923                 goto errout;
3924         }
3925         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3926         return;
3927 errout:
3928         if (err < 0)
3929                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3930 }
3931
3932 static inline size_t inet6_prefix_nlmsg_size(void)
3933 {
3934         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3935                + nla_total_size(sizeof(struct in6_addr))
3936                + nla_total_size(sizeof(struct prefix_cacheinfo));
3937 }
3938
3939 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3940                              struct prefix_info *pinfo, u32 pid, u32 seq,
3941                              int event, unsigned int flags)
3942 {
3943         struct prefixmsg *pmsg;
3944         struct nlmsghdr *nlh;
3945         struct prefix_cacheinfo ci;
3946
3947         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3948         if (nlh == NULL)
3949                 return -EMSGSIZE;
3950
3951         pmsg = nlmsg_data(nlh);
3952         pmsg->prefix_family = AF_INET6;
3953         pmsg->prefix_pad1 = 0;
3954         pmsg->prefix_pad2 = 0;
3955         pmsg->prefix_ifindex = idev->dev->ifindex;
3956         pmsg->prefix_len = pinfo->prefix_len;
3957         pmsg->prefix_type = pinfo->type;
3958         pmsg->prefix_pad3 = 0;
3959         pmsg->prefix_flags = 0;
3960         if (pinfo->onlink)
3961                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3962         if (pinfo->autoconf)
3963                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3964
3965         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3966
3967         ci.preferred_time = ntohl(pinfo->prefered);
3968         ci.valid_time = ntohl(pinfo->valid);
3969         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3970
3971         return nlmsg_end(skb, nlh);
3972
3973 nla_put_failure:
3974         nlmsg_cancel(skb, nlh);
3975         return -EMSGSIZE;
3976 }
3977
3978 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3979                          struct prefix_info *pinfo)
3980 {
3981         struct sk_buff *skb;
3982         struct net *net = dev_net(idev->dev);
3983         int err = -ENOBUFS;
3984
3985         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3986         if (skb == NULL)
3987                 goto errout;
3988
3989         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3990         if (err < 0) {
3991                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3992                 WARN_ON(err == -EMSGSIZE);
3993                 kfree_skb(skb);
3994                 goto errout;
3995         }
3996         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3997         return;
3998 errout:
3999         if (err < 0)
4000                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4001 }
4002
4003 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4004 {
4005         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4006
4007         switch (event) {
4008         case RTM_NEWADDR:
4009                 /*
4010                  * If the address was optimistic
4011                  * we inserted the route at the start of
4012                  * our DAD process, so we don't need
4013                  * to do it again
4014                  */
4015                 if (!(ifp->rt->rt6i_node))
4016                         ip6_ins_rt(ifp->rt);
4017                 if (ifp->idev->cnf.forwarding)
4018                         addrconf_join_anycast(ifp);
4019                 break;
4020         case RTM_DELADDR:
4021                 if (ifp->idev->cnf.forwarding)
4022                         addrconf_leave_anycast(ifp);
4023                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4024                 dst_hold(&ifp->rt->u.dst);
4025                 if (ip6_del_rt(ifp->rt))
4026                         dst_free(&ifp->rt->u.dst);
4027                 break;
4028         }
4029 }
4030
4031 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4032 {
4033         rcu_read_lock_bh();
4034         if (likely(ifp->idev->dead == 0))
4035                 __ipv6_ifa_notify(event, ifp);
4036         rcu_read_unlock_bh();
4037 }
4038
4039 #ifdef CONFIG_SYSCTL
4040
4041 static
4042 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4043                            void __user *buffer, size_t *lenp, loff_t *ppos)
4044 {
4045         int *valp = ctl->data;
4046         int val = *valp;
4047         loff_t pos = *ppos;
4048         int ret;
4049
4050         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4051
4052         if (write)
4053                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4054         if (ret)
4055                 *ppos = pos;
4056         return ret;
4057 }
4058
4059 static void dev_disable_change(struct inet6_dev *idev)
4060 {
4061         if (!idev || !idev->dev)
4062                 return;
4063
4064         if (idev->cnf.disable_ipv6)
4065                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4066         else
4067                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4068 }
4069
4070 static void addrconf_disable_change(struct net *net, __s32 newf)
4071 {
4072         struct net_device *dev;
4073         struct inet6_dev *idev;
4074
4075         rcu_read_lock();
4076         for_each_netdev_rcu(net, dev) {
4077                 idev = __in6_dev_get(dev);
4078                 if (idev) {
4079                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4080                         idev->cnf.disable_ipv6 = newf;
4081                         if (changed)
4082                                 dev_disable_change(idev);
4083                 }
4084         }
4085         rcu_read_unlock();
4086 }
4087
4088 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4089 {
4090         struct net *net;
4091
4092         net = (struct net *)table->extra2;
4093
4094         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4095                 return 0;
4096
4097         if (!rtnl_trylock()) {
4098                 /* Restore the original values before restarting */
4099                 *p = old;
4100                 return restart_syscall();
4101         }
4102
4103         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4104                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4105                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4106                 addrconf_disable_change(net, newf);
4107         } else if ((!*p) ^ (!old))
4108                 dev_disable_change((struct inet6_dev *)table->extra1);
4109
4110         rtnl_unlock();
4111         return 0;
4112 }
4113
4114 static
4115 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4116                             void __user *buffer, size_t *lenp, loff_t *ppos)
4117 {
4118         int *valp = ctl->data;
4119         int val = *valp;
4120         loff_t pos = *ppos;
4121         int ret;
4122
4123         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4124
4125         if (write)
4126                 ret = addrconf_disable_ipv6(ctl, valp, val);
4127         if (ret)
4128                 *ppos = pos;
4129         return ret;
4130 }
4131
4132 static struct addrconf_sysctl_table
4133 {
4134         struct ctl_table_header *sysctl_header;
4135         ctl_table addrconf_vars[DEVCONF_MAX+1];
4136         char *dev_name;
4137 } addrconf_sysctl __read_mostly = {
4138         .sysctl_header = NULL,
4139         .addrconf_vars = {
4140                 {
4141                         .procname       =       "forwarding",
4142                         .data           =       &ipv6_devconf.forwarding,
4143                         .maxlen         =       sizeof(int),
4144                         .mode           =       0644,
4145                         .proc_handler   =       addrconf_sysctl_forward,
4146                 },
4147                 {
4148                         .procname       =       "hop_limit",
4149                         .data           =       &ipv6_devconf.hop_limit,
4150                         .maxlen         =       sizeof(int),
4151                         .mode           =       0644,
4152                         .proc_handler   =       proc_dointvec,
4153                 },
4154                 {
4155                         .procname       =       "mtu",
4156                         .data           =       &ipv6_devconf.mtu6,
4157                         .maxlen         =       sizeof(int),
4158                         .mode           =       0644,
4159                         .proc_handler   =       proc_dointvec,
4160                 },
4161                 {
4162                         .procname       =       "accept_ra",
4163                         .data           =       &ipv6_devconf.accept_ra,
4164                         .maxlen         =       sizeof(int),
4165                         .mode           =       0644,
4166                         .proc_handler   =       proc_dointvec,
4167                 },
4168                 {
4169                         .procname       =       "accept_redirects",
4170                         .data           =       &ipv6_devconf.accept_redirects,
4171                         .maxlen         =       sizeof(int),
4172                         .mode           =       0644,
4173                         .proc_handler   =       proc_dointvec,
4174                 },
4175                 {
4176                         .procname       =       "autoconf",
4177                         .data           =       &ipv6_devconf.autoconf,
4178                         .maxlen         =       sizeof(int),
4179                         .mode           =       0644,
4180                         .proc_handler   =       proc_dointvec,
4181                 },
4182                 {
4183                         .procname       =       "dad_transmits",
4184                         .data           =       &ipv6_devconf.dad_transmits,
4185                         .maxlen         =       sizeof(int),
4186                         .mode           =       0644,
4187                         .proc_handler   =       proc_dointvec,
4188                 },
4189                 {
4190                         .procname       =       "router_solicitations",
4191                         .data           =       &ipv6_devconf.rtr_solicits,
4192                         .maxlen         =       sizeof(int),
4193                         .mode           =       0644,
4194                         .proc_handler   =       proc_dointvec,
4195                 },
4196                 {
4197                         .procname       =       "router_solicitation_interval",
4198                         .data           =       &ipv6_devconf.rtr_solicit_interval,
4199                         .maxlen         =       sizeof(int),
4200                         .mode           =       0644,
4201                         .proc_handler   =       proc_dointvec_jiffies,
4202                 },
4203                 {
4204                         .procname       =       "router_solicitation_delay",
4205                         .data           =       &ipv6_devconf.rtr_solicit_delay,
4206                         .maxlen         =       sizeof(int),
4207                         .mode           =       0644,
4208                         .proc_handler   =       proc_dointvec_jiffies,
4209                 },
4210                 {
4211                         .procname       =       "force_mld_version",
4212                         .data           =       &ipv6_devconf.force_mld_version,
4213                         .maxlen         =       sizeof(int),
4214                         .mode           =       0644,
4215                         .proc_handler   =       proc_dointvec,
4216                 },
4217 #ifdef CONFIG_IPV6_PRIVACY
4218                 {
4219                         .procname       =       "use_tempaddr",
4220                         .data           =       &ipv6_devconf.use_tempaddr,
4221                         .maxlen         =       sizeof(int),
4222                         .mode           =       0644,
4223                         .proc_handler   =       proc_dointvec,
4224                 },
4225                 {
4226                         .procname       =       "temp_valid_lft",
4227                         .data           =       &ipv6_devconf.temp_valid_lft,
4228                         .maxlen         =       sizeof(int),
4229                         .mode           =       0644,
4230                         .proc_handler   =       proc_dointvec,
4231                 },
4232                 {
4233                         .procname       =       "temp_prefered_lft",
4234                         .data           =       &ipv6_devconf.temp_prefered_lft,
4235                         .maxlen         =       sizeof(int),
4236                         .mode           =       0644,
4237                         .proc_handler   =       proc_dointvec,
4238                 },
4239                 {
4240                         .procname       =       "regen_max_retry",
4241                         .data           =       &ipv6_devconf.regen_max_retry,
4242                         .maxlen         =       sizeof(int),
4243                         .mode           =       0644,
4244                         .proc_handler   =       proc_dointvec,
4245                 },
4246                 {
4247                         .procname       =       "max_desync_factor",
4248                         .data           =       &ipv6_devconf.max_desync_factor,
4249                         .maxlen         =       sizeof(int),
4250                         .mode           =       0644,
4251                         .proc_handler   =       proc_dointvec,
4252                 },
4253 #endif
4254                 {
4255                         .procname       =       "max_addresses",
4256                         .data           =       &ipv6_devconf.max_addresses,
4257                         .maxlen         =       sizeof(int),
4258                         .mode           =       0644,
4259                         .proc_handler   =       proc_dointvec,
4260                 },
4261                 {
4262                         .procname       =       "accept_ra_defrtr",
4263                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4264                         .maxlen         =       sizeof(int),
4265                         .mode           =       0644,
4266                         .proc_handler   =       proc_dointvec,
4267                 },
4268                 {
4269                         .procname       =       "accept_ra_pinfo",
4270                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4271                         .maxlen         =       sizeof(int),
4272                         .mode           =       0644,
4273                         .proc_handler   =       proc_dointvec,
4274                 },
4275 #ifdef CONFIG_IPV6_ROUTER_PREF
4276                 {
4277                         .procname       =       "accept_ra_rtr_pref",
4278                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4279                         .maxlen         =       sizeof(int),
4280                         .mode           =       0644,
4281                         .proc_handler   =       proc_dointvec,
4282                 },
4283                 {
4284                         .procname       =       "router_probe_interval",
4285                         .data           =       &ipv6_devconf.rtr_probe_interval,
4286                         .maxlen         =       sizeof(int),
4287                         .mode           =       0644,
4288                         .proc_handler   =       proc_dointvec_jiffies,
4289                 },
4290 #ifdef CONFIG_IPV6_ROUTE_INFO
4291                 {
4292                         .procname       =       "accept_ra_rt_info_max_plen",
4293                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4294                         .maxlen         =       sizeof(int),
4295                         .mode           =       0644,
4296                         .proc_handler   =       proc_dointvec,
4297                 },
4298 #endif
4299 #endif
4300                 {
4301                         .procname       =       "proxy_ndp",
4302                         .data           =       &ipv6_devconf.proxy_ndp,
4303                         .maxlen         =       sizeof(int),
4304                         .mode           =       0644,
4305                         .proc_handler   =       proc_dointvec,
4306                 },
4307                 {
4308                         .procname       =       "accept_source_route",
4309                         .data           =       &ipv6_devconf.accept_source_route,
4310                         .maxlen         =       sizeof(int),
4311                         .mode           =       0644,
4312                         .proc_handler   =       proc_dointvec,
4313                 },
4314 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4315                 {
4316                         .procname       =       "optimistic_dad",
4317                         .data           =       &ipv6_devconf.optimistic_dad,
4318                         .maxlen         =       sizeof(int),
4319                         .mode           =       0644,
4320                         .proc_handler   =       proc_dointvec,
4321
4322                 },
4323 #endif
4324 #ifdef CONFIG_IPV6_MROUTE
4325                 {
4326                         .procname       =       "mc_forwarding",
4327                         .data           =       &ipv6_devconf.mc_forwarding,
4328                         .maxlen         =       sizeof(int),
4329                         .mode           =       0444,
4330                         .proc_handler   =       proc_dointvec,
4331                 },
4332 #endif
4333                 {
4334                         .procname       =       "disable_ipv6",
4335                         .data           =       &ipv6_devconf.disable_ipv6,
4336                         .maxlen         =       sizeof(int),
4337                         .mode           =       0644,
4338                         .proc_handler   =       addrconf_sysctl_disable,
4339                 },
4340                 {
4341                         .procname       =       "accept_dad",
4342                         .data           =       &ipv6_devconf.accept_dad,
4343                         .maxlen         =       sizeof(int),
4344                         .mode           =       0644,
4345                         .proc_handler   =       proc_dointvec,
4346                 },
4347                 {
4348                         .procname       = "force_tllao",
4349                         .data           = &ipv6_devconf.force_tllao,
4350                         .maxlen         = sizeof(int),
4351                         .mode           = 0644,
4352                         .proc_handler   = proc_dointvec
4353                 },
4354                 {
4355                         /* sentinel */
4356                 }
4357         },
4358 };
4359
4360 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4361                 struct inet6_dev *idev, struct ipv6_devconf *p)
4362 {
4363         int i;
4364         struct addrconf_sysctl_table *t;
4365
4366 #define ADDRCONF_CTL_PATH_DEV   3
4367
4368         struct ctl_path addrconf_ctl_path[] = {
4369                 { .procname = "net", },
4370                 { .procname = "ipv6", },
4371                 { .procname = "conf", },
4372                 { /* to be set */ },
4373                 { },
4374         };
4375
4376
4377         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4378         if (t == NULL)
4379                 goto out;
4380
4381         for (i=0; t->addrconf_vars[i].data; i++) {
4382                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4383                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4384                 t->addrconf_vars[i].extra2 = net;
4385         }
4386
4387         /*
4388          * Make a copy of dev_name, because '.procname' is regarded as const
4389          * by sysctl and we wouldn't want anyone to change it under our feet
4390          * (see SIOCSIFNAME).
4391          */
4392         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4393         if (!t->dev_name)
4394                 goto free;
4395
4396         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4397
4398         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4399                         t->addrconf_vars);
4400         if (t->sysctl_header == NULL)
4401                 goto free_procname;
4402
4403         p->sysctl = t;
4404         return 0;
4405
4406 free_procname:
4407         kfree(t->dev_name);
4408 free:
4409         kfree(t);
4410 out:
4411         return -ENOBUFS;
4412 }
4413
4414 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4415 {
4416         struct addrconf_sysctl_table *t;
4417
4418         if (p->sysctl == NULL)
4419                 return;
4420
4421         t = p->sysctl;
4422         p->sysctl = NULL;
4423         unregister_sysctl_table(t->sysctl_header);
4424         kfree(t->dev_name);
4425         kfree(t);
4426 }
4427
4428 static void addrconf_sysctl_register(struct inet6_dev *idev)
4429 {
4430         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4431                               &ndisc_ifinfo_sysctl_change);
4432         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4433                                         idev, &idev->cnf);
4434 }
4435
4436 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4437 {
4438         __addrconf_sysctl_unregister(&idev->cnf);
4439         neigh_sysctl_unregister(idev->nd_parms);
4440 }
4441
4442
4443 #endif
4444
4445 static int __net_init addrconf_init_net(struct net *net)
4446 {
4447         int err;
4448         struct ipv6_devconf *all, *dflt;
4449
4450         err = -ENOMEM;
4451         all = &ipv6_devconf;
4452         dflt = &ipv6_devconf_dflt;
4453
4454         if (!net_eq(net, &init_net)) {
4455                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4456                 if (all == NULL)
4457                         goto err_alloc_all;
4458
4459                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4460                 if (dflt == NULL)
4461                         goto err_alloc_dflt;
4462         } else {
4463                 /* these will be inherited by all namespaces */
4464                 dflt->autoconf = ipv6_defaults.autoconf;
4465                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4466         }
4467
4468         net->ipv6.devconf_all = all;
4469         net->ipv6.devconf_dflt = dflt;
4470
4471 #ifdef CONFIG_SYSCTL
4472         err = __addrconf_sysctl_register(net, "all", NULL, all);
4473         if (err < 0)
4474                 goto err_reg_all;
4475
4476         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4477         if (err < 0)
4478                 goto err_reg_dflt;
4479 #endif
4480         return 0;
4481
4482 #ifdef CONFIG_SYSCTL
4483 err_reg_dflt:
4484         __addrconf_sysctl_unregister(all);
4485 err_reg_all:
4486         kfree(dflt);
4487 #endif
4488 err_alloc_dflt:
4489         kfree(all);
4490 err_alloc_all:
4491         return err;
4492 }
4493
4494 static void __net_exit addrconf_exit_net(struct net *net)
4495 {
4496 #ifdef CONFIG_SYSCTL
4497         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4498         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4499 #endif
4500         if (!net_eq(net, &init_net)) {
4501                 kfree(net->ipv6.devconf_dflt);
4502                 kfree(net->ipv6.devconf_all);
4503         }
4504 }
4505
4506 static struct pernet_operations addrconf_ops = {
4507         .init = addrconf_init_net,
4508         .exit = addrconf_exit_net,
4509 };
4510
4511 /*
4512  *      Device notifier
4513  */
4514
4515 int register_inet6addr_notifier(struct notifier_block *nb)
4516 {
4517         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4518 }
4519
4520 EXPORT_SYMBOL(register_inet6addr_notifier);
4521
4522 int unregister_inet6addr_notifier(struct notifier_block *nb)
4523 {
4524         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4525 }
4526
4527 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4528
4529 /*
4530  *      Init / cleanup code
4531  */
4532
4533 int __init addrconf_init(void)
4534 {
4535         int err;
4536
4537         if ((err = ipv6_addr_label_init()) < 0) {
4538                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4539                         err);
4540                 return err;
4541         }
4542
4543         register_pernet_subsys(&addrconf_ops);
4544
4545         /* The addrconf netdev notifier requires that loopback_dev
4546          * has it's ipv6 private information allocated and setup
4547          * before it can bring up and give link-local addresses
4548          * to other devices which are up.
4549          *
4550          * Unfortunately, loopback_dev is not necessarily the first
4551          * entry in the global dev_base list of net devices.  In fact,
4552          * it is likely to be the very last entry on that list.
4553          * So this causes the notifier registry below to try and
4554          * give link-local addresses to all devices besides loopback_dev
4555          * first, then loopback_dev, which cases all the non-loopback_dev
4556          * devices to fail to get a link-local address.
4557          *
4558          * So, as a temporary fix, allocate the ipv6 structure for
4559          * loopback_dev first by hand.
4560          * Longer term, all of the dependencies ipv6 has upon the loopback
4561          * device and it being up should be removed.
4562          */
4563         rtnl_lock();
4564         if (!ipv6_add_dev(init_net.loopback_dev))
4565                 err = -ENOMEM;
4566         rtnl_unlock();
4567         if (err)
4568                 goto errlo;
4569
4570         register_netdevice_notifier(&ipv6_dev_notf);
4571
4572         addrconf_verify(0);
4573
4574         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4575         if (err < 0)
4576                 goto errout;
4577
4578         /* Only the first call to __rtnl_register can fail */
4579         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4580         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4581         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4582         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4583         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4584
4585         ipv6_addr_label_rtnl_register();
4586
4587         return 0;
4588 errout:
4589         unregister_netdevice_notifier(&ipv6_dev_notf);
4590 errlo:
4591         unregister_pernet_subsys(&addrconf_ops);
4592
4593         return err;
4594 }
4595
4596 void addrconf_cleanup(void)
4597 {
4598         struct inet6_ifaddr *ifa;
4599         struct net_device *dev;
4600         int i;
4601
4602         unregister_netdevice_notifier(&ipv6_dev_notf);
4603         unregister_pernet_subsys(&addrconf_ops);
4604
4605         rtnl_lock();
4606
4607         /* clean dev list */
4608         for_each_netdev(&init_net, dev) {
4609                 if (__in6_dev_get(dev) == NULL)
4610                         continue;
4611                 addrconf_ifdown(dev, 1);
4612         }
4613         addrconf_ifdown(init_net.loopback_dev, 2);
4614
4615         /*
4616          *      Check hash table.
4617          */
4618         write_lock_bh(&addrconf_hash_lock);
4619         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4620                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4621                         struct inet6_ifaddr *bifa;
4622
4623                         bifa = ifa;
4624                         ifa = ifa->lst_next;
4625                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4626                         /* Do not free it; something is wrong.
4627                            Now we can investigate it with debugger.
4628                          */
4629                 }
4630         }
4631         write_unlock_bh(&addrconf_hash_lock);
4632
4633         del_timer(&addr_chk_timer);
4634         rtnl_unlock();
4635 }